-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
5680 lines (4930 loc) · 469 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<html lang="en" style="scroll-padding-top: 70px;">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0, shrink-to-fit=no">
<link rel="stylesheet"
href="https://fonts.googleapis.com/css?family=Open+Sans:300italic,400italic,600italic,700italic,800italic,400,300,600,700,800">
<link rel="stylesheet" href="https://fonts.googleapis.com/css?family=Lora:400,700,400italic,700italic">
<link href="https://fonts.googleapis.com/css2?family=Exo:wght@400;700&family=Lato:wght@400;700&display=swap" rel="stylesheet">
<link rel="stylesheet" href="/static/expo/fonts/font-awesome.min.css">
<link rel="stylesheet" href="https://use.fontawesome.com/releases/v5.8.1/css/all.css" integrity="sha384-50oBUHEmvpQ+1lW4y57PTFmhCaXp0ML5d60M1M7uH2+nqUivzIebhndOJK28anvf" crossorigin="anonymous">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/bootstrap-select@1.13.14/dist/css/bootstrap-select.min.css">
<link rel="stylesheet" href="cards.css">
<link rel="stylesheet" href="https://use.fontawesome.com/releases/v5.8.1/css/all.css" integrity="sha384-50oBUHEmvpQ+1lW4y57PTFmhCaXp0ML5d60M1M7uH2+nqUivzIebhndOJK28anvf" crossorigin="anonymous">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/bootstrap@4.6.2/dist/css/bootstrap.min.css" integrity="sha384-xOolHFLEh07PJGoPkLv1IbcEPTNtaed2xpHsD9ESMhqIYd0nLMwNLD69Npy4HI+N" crossorigin="anonymous">
<script src="https://code.jquery.com/jquery-3.6.1.min.js"
integrity="sha256-o88AwQnZB+VDvE9tvIXrMQaPlFFSUTR+nldQm1LuPXQ=" crossorigin="anonymous"></script>
</script>
<script>
if (typeof jQuery === 'undefined') {
var script = document.createElement('script');
script.type = 'text/javascript';
script.src = "/static/core/js/jquery-3.6.1.min.js";
document.head.appendChild(script);
}
</script>
<script src="https://d3js.org/d3.v5.min.js"></script>
<script src="https://cdn.jsdelivr.net/npm/popper.js@1.16.0/dist/umd/popper.min.js" integrity="sha384-Q6E9RHvbIyZFJoft+2mJbHaEWldlvI9IOYy5n3zV9zzTtmI3UksdQRVvoxMfooAo" crossorigin="anonymous"></script>
<script src="https://cdn.jsdelivr.net/npm/bootstrap@4.6.2/dist/js/bootstrap.bundle.min.js" integrity="sha384-Fy6S3B9q64WdZWQUiU+q4/2Lc9npb8tCaSX9FK7E8HnRr0Jz8D6OP9dO5Vg3Q9ct" crossorigin="anonymous"></script>
<script src="https://cdn.jsdelivr.net/npm/bootstrap-select@1.13.14/dist/js/bootstrap-select.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/corejs-typeahead/1.3.1/typeahead.bundle.min.js" integrity="sha512-lEb9Vp/rkl9g2E/LdHIMFTqz21+LA79f84gqP75fbimHqVTu6483JG1AwJlWLLQ8ezTehty78fObKupq3HSHPQ==" crossorigin="anonymous"></script>
<script src="https://cdn.jsdelivr.net/npm/moment@2.24.0/min/moment.min.js"
integrity="sha256-4iQZ6BVL4qNKlQ27TExEhBN1HFPvAvAMbFavKKosSWQ="
crossorigin="anonymous"></script>
<script src="https://cdn.jsdelivr.net/npm/js-cookie@2/src/js.cookie.min.js"></script>
<script src="/static/core/js/ajax-csrf-snippet.js" type="text/javascript"></script>
<script src="/static/virtual/js/virtual.js"></script>
<link rel="stylesheet" href="virtual.css">
<style>
body {
background: #f6f6f6;
}
</style>
</head>
<body>
<!-- NAV -->
<!--
<nav class="navbar sticky-top navbar-expand-lg navbar-light mr-auto" id="main-nav">
<div class="container-fluid">
<a class="navbar-brand" href="/">
<img src="/static/core/img/ICML-logo.svg" height="40px">
</a>
<button class="navbar-toggler" type="button" data-toggle="collapse" data-target="#navbarNav"
aria-controls="navbarNav" aria-expanded="false" aria-label="Toggle navigation">
<span class="navbar-toggler-icon"></span>
</button>
<div class="collapse navbar-collapse text-right flex-grow-1" id="navbarNav">
<ul class="navbar-nav ml-auto">
<li class="nav-item ">
<a class="nav-link" href="/virtual/2022/events/tutorial">Tutorials</a>
</li>
<li class="nav-item dropdown">
<a class="nav-link dropdown-toggle" href="#" id="navbarDropdown" role="button"
data-toggle="dropdown" aria-haspopup="true" aria-expanded="false">
Main Conference
</a>
<div class="dropdown-menu" aria-labelledby="navbarDropdown">
<a class="dropdown-item" href="/virtual/2022/events/oral">Orals</a>
<div class="dropdown-divider"></div>
<a class="dropdown-item" href="/virtual/2022/papers.html">Papers</a>
</div>
</li>
</ul>
</div>
</div>
</nav>
-->
<!-- NAV -->
<nav class="navbar sticky-top navbar-expand-lg navbar-light mr-auto" id="main-nav">
<div class="container-fluid">
<a class="navbar-brand" href="">
<img src="tmlr_logo.jpeg" height="40px">
Transactions on Machine Learning Research
</a>
<button class="navbar-toggler" type="button" data-toggle="collapse" data-target="#navbarNav"
aria-controls="navbarNav" aria-expanded="false" aria-label="Toggle navigation">
<span class="navbar-toggler-icon"></span>
</button>
<div class="collapse navbar-collapse text-right flex-grow-1" id="navbarNav">
<ul class="navbar-nav ml-auto">
<!--
<li class="nav-item dropdown">
<a class="nav-link dropdown-toggle" href="#" id="navbarDropdown" role="button"
data-toggle="dropdown" aria-haspopup="true" aria-expanded="false">
Main Conference
</a>
<div class="dropdown-menu" aria-labelledby="navbarDropdown">
<div class="dropdown-divider"></div>
<a class="dropdown-item" href="/virtual/2023/events/oral">Orals</a>
<div class="dropdown-divider"></div>
<a class="dropdown-item" href="/virtual/2023/events/spotlight">Spotlights</a>
<div class="dropdown-divider"></div>
<a class="dropdown-item" href="/virtual/2023/papers.html">Papers</a>
</div>
</li>
-->
<!--
<li class="nav-item">
<a class="nav-link" href="../">All Papers</a>
</li>
-->
<!--
<li class="nav-item">
<a class="nav-link" href="">Papers with Videos</a>
</li>
-->
<!--
<li class="nav-item">
<a class="nav-link" href="../featured_papers.html">Featured Papers</a>
</li>
-->
<!--
<li class="nav-item ">
<a class="nav-link" href="/virtual/2023/search"><i class="fas fa-search"></i> </a>
</li>
-->
</ul>
</div>
</div>
</nav>
<div class="container">
<br />
<div class="row">
<div class="col-md-12"></div>
<div class="title-centered" style="text-align:center">TMLR Infinite Conference</div>
</div>
</div>
<div class="row">
<div class="col-sm-12">
<div style="max-width: 1500px; margin:auto; border">
<div class="grid-displaycards">
<div class="displaycards touchup-date" id="event-lIy0TEUou7">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/lIy0TEUou7.html">Modular Quantization-Aware Training for 6D Object Pose Estimation</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Saqib Javed · Chengkun Li · Andrew Lawrence Price · Yinlin Hu · Mathieu Salzmann</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-lIy0TEUou7"></div>
<a href="paper_pages/lIy0TEUou7.html">
<img src="http://img.youtube.com/vi/EBNr0qNem8U/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-lIy0TEUou7" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-lIy0TEUou7" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-lIy0TEUou7">
Abstract <i id="caret-lIy0TEUou7" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-lIy0TEUou7">
<div class="abstract-display">
<p>Edge applications, such as collaborative robotics and spacecraft rendezvous, demand efficient 6D object pose estimation on resource-constrained embedded platforms. Existing 6D object pose estimation networks are often too large for such deployments, necessitating compression while maintaining reliable performance. To address this challenge, we introduce Modular Quantization-Aware Training (MQAT), an adaptive and mixed-precision quantization-aware training strategy that exploits the modular structure of modern 6D object pose estimation architectures. MQAT guides a systematic gradated modular quantization sequence and determines module-specific bit precisions, leading to quantized models that outperform those produced by state-of-the-art uniform and mixed-precision quantization techniques. Our experiments showcase the generality of MQAT across datasets, architectures, and quantization algorithms. Additionally, we observe that MQAT quantized models can achieve an accuracy boost (>7% ADI-0.1d) over the baseline full-precision network while reducing model size by a factor of 4x or more.
Project Page: https://saqibjaved1.github.io/MQAT_</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-lh6vOAHuvo">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/lh6vOAHuvo.html">AGaLiTe: Approximate Gated Linear Transformers for Online Reinforcement Learning</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Subhojeet Pramanik · Esraa Elelimy · Marlos C. Machado · Adam White</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-lh6vOAHuvo"></div>
<a href="paper_pages/lh6vOAHuvo.html">
<img src="http://img.youtube.com/vi/-bTe48JIUds/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-lh6vOAHuvo" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-lh6vOAHuvo" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-lh6vOAHuvo">
Abstract <i id="caret-lh6vOAHuvo" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-lh6vOAHuvo">
<div class="abstract-display">
<p>In this paper we investigate transformer architectures designed for partially observable online reinforcement learning. The self-attention mechanism in the transformer architecture is capable of capturing long-range dependencies and it is the main reason behind its effectiveness in processing sequential data. Nevertheless, despite their success, transformers have two significant drawbacks that still limit their applicability in online reinforcement learning: (1) in order to remember all past information, the self-attention mechanism requires access to the whole history to be provided as context. (2) The inference cost in transformers is expensive. In this paper, we introduce recurrent alternatives to the transformer self-attention mechanism that offer context-independent inference cost, leverage long-range dependencies effectively, and performs well in online reinforcement learning task. We quantify the impact of the different components of our architecture in a diagnostic environment and assess performance gains in 2D and 3D pixel-based partially-observable environments (e.g. T-Maze, Mystery Path, Craftax, and Memory Maze). Compared with a state-of-the-art architecture, GTrXL, inference in our approach is at least 40% cheaper while reducing memory use more than 50%. Our approach either performs similarly or better than GTrXL, improving more than 37% upon GTrXL performance in harder tasks.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-0uwe0z2Hqm">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/0uwe0z2Hqm.html">Deep-Graph-Sprints: Accelerated Representation Learning in Continuous-Time Dynamic Graphs</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Ahmad Naser Eddin · Jacopo Bono · David Oliveira Aparicio · Hugo Ferreira · Pedro Manuel Pinto Ribeiro · Pedro Bizarro</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-0uwe0z2Hqm"></div>
<a href="paper_pages/0uwe0z2Hqm.html">
<img src="http://img.youtube.com/vi/PL6HOcHf_w6mrceEUlo0TgW7jioxXV38o6/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-0uwe0z2Hqm" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-0uwe0z2Hqm" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-0uwe0z2Hqm">
Abstract <i id="caret-0uwe0z2Hqm" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-0uwe0z2Hqm">
<div class="abstract-display">
<p>Continuous-time dynamic graphs (CTDGs) are essential for modeling interconnected, evolving systems. Traditional methods for extracting knowledge from these graphs often depend on feature engineering or deep learning. Feature engineering is limited by the manual and time-intensive nature of crafting features, while deep learning approaches suffer from high inference latency, making them impractical for real-time applications. This paper introduces Deep-Graph-Sprints (DGS), a novel deep learning architecture designed for efficient representation learning on CTDGs with low-latency inference requirements. We benchmark DGS against state-of-the-art (SOTA) feature engineering and graph neural network methods using five diverse datasets. The results indicate that DGS achieves competitive performance while inference speed improves between 4x and 12x compared to other deep learning approaches on our benchmark datasets. Our method effectively bridges the gap between deep representation learning and low-latency application requirements for CTDGs.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-fJEsas1z8J">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/fJEsas1z8J.html">MoCaE: Mixture of Calibrated Experts Significantly Improves Object Detection</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Kemal Oksuz · Selim Kuzucu · Tom Joy · Puneet K. Dokania</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-fJEsas1z8J"></div>
<a href="paper_pages/fJEsas1z8J.html">
<img src="http://img.youtube.com/vi/PzyXl5VBrqE/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-fJEsas1z8J" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-fJEsas1z8J" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-fJEsas1z8J">
Abstract <i id="caret-fJEsas1z8J" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-fJEsas1z8J">
<div class="abstract-display">
<p>Combining the strengths of many existing predictors to obtain a Mixture of Experts which is superior to its individual components is an effective way to improve the performance without having to develop new architectures or train a model from scratch. However, surprisingly, we find that naively combining off-the-shelf object detectors in a similar way to Deep Ensembles, can often lead to degraded performance. We identify that the primary cause of this issue is that the predictions of the experts do not match their performance, a term referred to as miscalibration. Consequently, the most confident detector dominates the final predictions, preventing the mixture from leveraging all the predictions from the experts appropriately. To address this, when constructing the Mixture of Experts for object detection, we propose to combine their predictions in a manner which reflects the individual performance of the experts; an objective we achieve by first calibrating the predictions before filtering and refining them. We term this approach the Mixture of Calibrated Experts (MoCaE) and demonstrate its effectiveness through extensive experiments on 5 different detection tasks, showing that it: (i) improves object detectors on COCO and instance segmentation methods on LVIS by up to $\sim 2.5$ AP; (ii) reaches state-of-the-art on COCO test-dev with $65.1$ AP and on DOTA with $82.62$ $\mathrm{AP_{50}}$; (iii) outperforms single models consistently on recent detection tasks such as Open Vocabulary Object Detection. Code is available at: https://github.com/fiveai/MoCaE</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-6j5M75iK3a">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/6j5M75iK3a.html">Continual Learning in Open-vocabulary Classification with Complementary Memory Systems</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Zhen Zhu · Weijie Lyu · Yao Xiao · Derek Hoiem</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-6j5M75iK3a"></div>
<a href="paper_pages/6j5M75iK3a.html">
<img src="http://img.youtube.com/vi/RkmeJupwNkY/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-6j5M75iK3a" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-6j5M75iK3a" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-6j5M75iK3a">
Abstract <i id="caret-6j5M75iK3a" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-6j5M75iK3a">
<div class="abstract-display">
<p>We introduce a method for flexible and efficient continual learning in open-vocabulary image classification, drawing inspiration from the complementary learning systems observed in human cognition. Specifically, we propose to combine predictions from a CLIP zero-shot model and the exemplar-based model, using the zero-shot estimated probability that a sample's class is within the exemplar classes. We also propose a ``tree probe'' method, an adaption of lazy learning principles, which enables fast learning from new examples with competitive accuracy to batch-trained linear models. We test in data incremental, class incremental, and task incremental settings, as well as ability to perform flexible inference on varying subsets of zero-shot and learned categories. Our proposed method achieves a good balance of learning speed, target task effectiveness, and zero-shot effectiveness. Code is available at https://github.com/jessemelpolio/TreeProbe.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-mDGvrH7lju">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/mDGvrH7lju.html">CFASL: Composite Factor-Aligned Symmetry Learning for Disentanglement in Variational AutoEncoder</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Hee-Jun Jung · Jaehyoung Jeong · Kangil Kim</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-mDGvrH7lju"></div>
<a href="paper_pages/mDGvrH7lju.html">
<img src="http://img.youtube.com/vi/R03AoD3SRZ8/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-mDGvrH7lju" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-mDGvrH7lju" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-mDGvrH7lju">
Abstract <i id="caret-mDGvrH7lju" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-mDGvrH7lju">
<div class="abstract-display">
<p>Symmetries of input and latent vectors have provided valuable insights for disentanglement learning in VAEs. However, only a few works were proposed as an unsupervised method, and even these works require known factor information in training data. We propose a
novel method, Composite Factor-Aligned Symmetry Learning (CFASL), which is integrated into VAEs for learning symmetry-based disentanglement in unsupervised learning without any knowledge of the dataset factor information. CFASL incorporates three novel features for learning symmetry-based disentanglement: 1) Injecting inductive bias to align latent vector dimensions to factor-aligned symmetries within an explicit learnable symmetry code-book 2) Learning a composite symmetry to express unknown factors change between two random samples by learning factor-aligned symmetries within the codebook 3) Inducing group equivariant encoder and decoder in training VAEs with the two conditions. In addition, we propose an extended evaluation metric for multi-factor changes in comparison to disentanglement evaluation in VAEs. In quantitative and in-depth qualitative analysis, CFASL demonstrates a significant improvement of disentanglement in single-factor change, and multi-factor change conditions compared to state-of-the-art methods.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-3YlOr7BHkx">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/3YlOr7BHkx.html">Mislabeled examples detection viewed as probing machine learning models: concepts, survey and extensive benchmark</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Thomas George · Pierre Nodet · Alexis Bondu · Vincent Lemaire</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-3YlOr7BHkx"></div>
<a href="paper_pages/3YlOr7BHkx.html">
<img src="http://img.youtube.com/vi/fT9VZXs0nh8/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-3YlOr7BHkx" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-3YlOr7BHkx" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-3YlOr7BHkx">
Abstract <i id="caret-3YlOr7BHkx" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-3YlOr7BHkx">
<div class="abstract-display">
<p>Mislabeled examples are ubiquitous in real-world machine learning datasets, advocating the development of techniques for automatic detection. We show that most mislabeled detection methods can be viewed as probing trained machine learning models using a few core principles. We formalize a modular framework that encompasses these methods, parameterized by only 4 building blocks, as well as a Python library that demonstrates that these principles can actually be implemented. The focus is on classifier-agnostic concepts, with an emphasis on adapting methods developed for deep learning models to non-deep classifiers for tabular data. We benchmark existing methods on (artificial) Completely At Random (NCAR) as well as (realistic) Not At Random (NNAR) labeling noise from a variety of tasks with imperfect labeling rules. This benchmark provides new insights as well as limitations of existing methods in this setup.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-2noXK5KBbx">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/2noXK5KBbx.html">Graph Structure Learning with Interpretable Bayesian Neural Networks</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Max Wasserman · Gonzalo Mateos</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-2noXK5KBbx"></div>
<a href="paper_pages/2noXK5KBbx.html">
<img src="http://img.youtube.com/vi/zcYD-r8DlUI/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-2noXK5KBbx" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-2noXK5KBbx" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-2noXK5KBbx">
Abstract <i id="caret-2noXK5KBbx" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-2noXK5KBbx">
<div class="abstract-display">
<p>Graphs serve as generic tools to encode the underlying relational structure of data. Often this graph is not given, and so the task of inferring it from nodal observations becomes important. Traditional approaches formulate a convex inverse problem with a smoothness promoting objective and rely on iterative methods to obtain a solution. In supervised settings where graph labels are available, one can unroll and truncate these iterations into a deep network that is trained end-to-end. Such a network is parameter efficient and inherits inductive bias from the optimization formulation, an appealing aspect for data constrained settings in, e.g., medicine, finance, and the natural sciences. But typically such settings care equally about \textit{uncertainty} over edge predictions, not just point estimates. Here we introduce novel iterations with \textit{independently interpretable parameters}, i.e., parameters whose values - independent of other parameters' settings - proportionally influence characteristics of the estimated graph, such as edge sparsity. After unrolling these iterations, prior knowledge over such graph characteristics shape \textit{prior distributions} over these independently interpretable network parameters to yield a Bayesian neural network (BNN) capable of graph structure learning (GSL) from smooth signal observations. Fast execution and parameter efficiency allow for high-fidelity posterior approximation via Markov Chain Monte Carlo (MCMC) and thus uncertainty quantification on edge predictions. Informative priors unlock modeling tools from Bayesian statistics like prior predictive checks. Synthetic and real data experiments corroborate this model's ability to provide well-calibrated estimates of uncertainty, in test cases that include unveiling economic sector modular structure from S$\&$P$500$ data and recovering pairwise digit similarities from MNIST images. Overall, this framework enables GSL in modest-scale applications where uncertainty on the data structure is paramount.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-hrKHkmLUFk">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/hrKHkmLUFk.html">Multi-intention Inverse Q-learning for Interpretable Behavior Representation</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Hao Zhu · Brice De La Crompe · Gabriel Kalweit · Artur Schneider · Maria Kalweit · Ilka Diester · Joschka Boedecker</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-hrKHkmLUFk"></div>
<a href="paper_pages/hrKHkmLUFk.html">
<img src="http://img.youtube.com/vi/0u-fboAO6-I/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-hrKHkmLUFk" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-hrKHkmLUFk" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-hrKHkmLUFk">
Abstract <i id="caret-hrKHkmLUFk" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-hrKHkmLUFk">
<div class="abstract-display">
<p>In advancing the understanding of natural decision-making processes, inverse reinforcement learning (IRL) methods have proven instrumental in reconstructing animal's intentions underlying complex behaviors. Given the recent development of a continuous-time multi-intention IRL framework, there has been persistent inquiry into inferring discrete time-varying rewards with IRL. To address this challenge, we introduce the class of hierarchical inverse Q-learning (HIQL) algorithms. Through an unsupervised learning process, HIQL divides expert trajectories into multiple intention segments, and solves the IRL problem independently for each. Applying HIQL to simulated experiments and several real animal behavior datasets, our approach outperforms current benchmarks in behavior prediction and produces interpretable reward functions. Our results suggest that the intention transition dynamics underlying complex decision-making behavior is better modeled by a step function instead of a smoothly varying function. This advancement holds promise for neuroscience and cognitive science, contributing to a deeper understanding of decision-making and uncovering underlying brain mechanisms.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-kUuPUIPvJ6">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/kUuPUIPvJ6.html">Support-Set Context Matters for Bongard Problems</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Nikhil Raghuraman · Adam W Harley · Leonidas Guibas</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-kUuPUIPvJ6"></div>
<a href="paper_pages/kUuPUIPvJ6.html">
<img src="http://img.youtube.com/vi/JO00GQHp0mQ/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-kUuPUIPvJ6" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-kUuPUIPvJ6" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-kUuPUIPvJ6">
Abstract <i id="caret-kUuPUIPvJ6" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-kUuPUIPvJ6">
<div class="abstract-display">
<p>Current machine learning methods struggle to solve Bongard problems, which are a type of IQ test that requires deriving an abstract “concept” from a set of positive and negative “support” images, and then classifying whether or not a new query image depicts the key concept. On Bongard-HOI, a benchmark for natural-image Bongard problems, most existing methods have reached at best 69% accuracy (where chance is 50%). Low accuracy is often attributed to neural nets’ lack of ability to find human-like symbolic rules. In this work, we point out that many existing methods are forfeiting accuracy due to a much simpler problem: they do not adapt image features given information contained in the support set as a whole, and rely instead on information extracted from individual supports. This is a critical issue, because the “key concept” in a typical Bongard problem can often only be distinguished using multiple positives and multiple negatives. We explore simple methods to incorporate this context and show substantial gains over prior works, leading to new state-of-the-art accuracy on Bongard-LOGO (75.3%) and Bongard-HOI (76.4%) compared to methods with equivalent vision backbone architectures and strong performance on the original Bongard problem set (60.8%). Code is available at https://github.com/nraghuraman/bongard-context.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-llQXLfbGOq">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/llQXLfbGOq.html">Attention Normalization Impacts Cardinality Generalization in Slot Attention</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Markus Krimmel · Jan Achterhold · Joerg Stueckler</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-llQXLfbGOq"></div>
<a href="paper_pages/llQXLfbGOq.html">
<img src="http://img.youtube.com/vi/dnTrHyZgyCY/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-llQXLfbGOq" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-llQXLfbGOq" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-llQXLfbGOq">
Abstract <i id="caret-llQXLfbGOq" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-llQXLfbGOq">
<div class="abstract-display">
<p>Object-centric scene decompositions are important representations for downstream tasks in fields such as computer vision and robotics. The recently proposed Slot Attention module, already leveraged by several derivative works for image segmentation and object tracking in videos, is a deep learning component which performs unsupervised object-centric scene decomposition on input images. It is based on an attention architecture, in which latent slot vectors, which hold compressed information on objects, attend to localized perceptual features from the input image. In this paper, we demonstrate that design decisions on normalizing the aggregated values in the attention architecture have considerable impact on the capabilities of Slot Attention to generalize to a higher number of slots and objects as seen during training. We propose and investigate alternatives to the original normalization scheme which increase the generalization capabilities of Slot Attention to varying slot and object counts, resulting in performance gains on the task of unsupervised image segmentation. The newly proposed normalizations represent minimal and easy to implement modifications of the usual Slot Attention module, changing the value aggregation mechanism from a weighted mean operation to a scaled weighted sum operation.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-2D36otXvBE">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/2D36otXvBE.html">Identifying and Clustering Counter Relationships of Team Compositions in PvP Games for Efficient Balance Analysis</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Chiu-Chou Lin · Yu-Wei Shih · Kuei-Ting Kuo · Yu-Cheng Chen · Chien-Hua Chen · Wei-Chen Chiu · I-Chen Wu</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-2D36otXvBE"></div>
<a href="paper_pages/2D36otXvBE.html">
<img src="http://img.youtube.com/vi/CmgoUtRqmI8/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-2D36otXvBE" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-2D36otXvBE" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-2D36otXvBE">
Abstract <i id="caret-2D36otXvBE" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-2D36otXvBE">
<div class="abstract-display">
<p>\textbf{How can balance be quantified in game settings?} This question is crucial for game designers, especially in player-versus-player (PvP) games, where analyzing the strength relations among predefined team compositions—such as hero combinations in multiplayer online battle arena (MOBA) games or decks in card games—is essential for enhancing gameplay and achieving balance. We have developed two advanced measures that extend beyond the simplistic win rate to quantify balance in zero-sum competitive scenarios. These measures are derived from win value estimations, which employ strength rating approximations via the Bradley-Terry model and counter relationship approximations via vector quantization, significantly reducing the computational complexity associated with traditional win value estimations. Throughout the learning process of these models, we identify useful categories of compositions and pinpoint their counter relationships, aligning with the experiences of human players without requiring specific game knowledge. Our methodology hinges on a simple technique to enhance codebook utilization in discrete representation with a deterministic vector quantization process for an extremely small state space. Our framework has been validated in popular online games, including \textit{Age of Empires II}, \textit{Hearthstone}, \textit{Brawl Stars}, and \textit{League of Legends}. The accuracy of the observed strength relations in these games is comparable to traditional pairwise win value predictions, while also offering a more manageable complexity for analysis. Ultimately, our findings contribute to a deeper understanding of PvP game dynamics and present a methodology that significantly improves game balance evaluation and design.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-IEKtMMSblm">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/IEKtMMSblm.html">PLUM: Improving Inference Efficiency By Leveraging Repetition-Sparsity Trade-Off</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Sachit Kuhar · Yash Jain · Alexey Tumanov</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-IEKtMMSblm"></div>
<a href="paper_pages/IEKtMMSblm.html">
<img src="http://img.youtube.com/vi/nE_CYDWqQ_I/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-IEKtMMSblm" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-IEKtMMSblm" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-IEKtMMSblm">
Abstract <i id="caret-IEKtMMSblm" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-IEKtMMSblm">
<div class="abstract-display">
<p>Efficient inference of Deep Neural Networks (DNNs) on resource-constrained edge devices is essential. Quantization and sparsity are key techniques that translate to repetition and sparsity within tensors at the hardware-software interface. This paper introduces the concept of repetition-sparsity trade-off that helps explain computational efficiency during inference. We propose PLUM, a unified co-design framework that integrates DNN inference systems and quantization (forward and backward pass) to leverage the repetition-sparsity trade-off to improve inference efficiency. Our results demonstrate that PLUM’s quantization method is more accurate than binary quantization with the same number of non-zero weights. Detailed analysis indicates that signed binarization generates a smaller distribution of effectual (non-zero) parameters nested within a larger distribution of total parameters of latent full-precision weights for a DNN block. Finally, the proposed PLUM framework achieves a 26% speedup on real hardware, doubles energy efficiency, and reduces density by 2.8× compared to binary methods while retaining top-1 accuracy when compared to prior-art methods for ResNets on ImageNet (by achieving 66.2% top-1 accuracy), presenting an alternative solution for deploying efficient models in resource-limited environments</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-1WqLLYgBNt">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/1WqLLYgBNt.html">Do not trust what you trust: Miscalibration in Semisupervised Learning</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Shambhavi Mishra · Balamurali Murugesan · Ismail Ben Ayed · Marco Pedersoli · Jose Dolz</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-1WqLLYgBNt"></div>
<a href="paper_pages/1WqLLYgBNt.html">
<img src="http://img.youtube.com/vi/Gj3-NgXo9Wk/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-1WqLLYgBNt" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-1WqLLYgBNt" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-1WqLLYgBNt">
Abstract <i id="caret-1WqLLYgBNt" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-1WqLLYgBNt">
<div class="abstract-display">
<p>State-of-the-art semi-supervised learning (SSL) approaches rely on highly confident predictions to serve as pseudo-labels that guide the training on unlabeled samples. An inherent drawback of this strategy stems from the quality of the uncertainty estimates, as pseudo-labels are filtered only based on their degree of uncertainty, regardless of the correctness of their predictions. Thus, assessing and enhancing the uncertainty of network predictions is of paramount importance in the pseudo-labeling process. In this work, we empirically demonstrate that SSL methods based on pseudo-labels are significantly miscalibrated, and formally demonstrate the minimization of the min-entropy, a lower bound of the Shannon entropy, as a potential cause for miscalibration. To alleviate this issue, we integrate a simple penalty term, which enforces the logit distances of the predictions on unlabeled samples to remain low, preventing the network predictions to become overconfident. Comprehensive experiments on a variety of SSL image classification benchmarks demonstrate that the proposed solution systematically improves the calibration performance of relevant SSL models, while also enhancing their discriminative power, being an appealing addition to tackle SSL tasks.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-TB18G0w6Ld">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/TB18G0w6Ld.html">Fairness Under Demographic Scarce Regime</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Patrik Joslin Kenfack · Samira Ebrahimi Kahou · Ulrich Aïvodji</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-TB18G0w6Ld"></div>
<a href="paper_pages/TB18G0w6Ld.html">
<img src="http://img.youtube.com/vi/dLMC1IbR4LA/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-TB18G0w6Ld" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-TB18G0w6Ld" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-TB18G0w6Ld">
Abstract <i id="caret-TB18G0w6Ld" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-TB18G0w6Ld">
<div class="abstract-display">
<p>Most existing works on fairness assume the model has full access to demographic information. However, there exist scenarios where demographic information is partially available because a record was not maintained throughout data collection or for privacy reasons. This setting is known as demographic scarce regime. Prior research has shown that training an attribute classifier to replace the missing sensitive attributes (proxy) can still improve fairness. However, using proxy-sensitive attributes worsens fairness-accuracy tradeoffs compared to true sensitive attributes. To address this limitation, we propose a framework to build attribute classifiers that achieve better fairness-accuracy tradeoffs. Our method introduces uncertainty awareness in the attribute classifier and enforces fairness on samples with demographic information inferred with the lowest uncertainty. We show empirically that enforcing fairness constraints on samples with uncertain sensitive attributes can negatively impact the fairness-accuracy tradeoff. Our experiments on five datasets showed that the proposed framework yields models with significantly better fairness-accuracy tradeoffs than classic attribute classifiers. Surprisingly, our framework can outperform models trained with fairness constraints on the true sensitive attributes in most benchmarks. We also show that these findings are consistent with other uncertainty measures such as conformal prediction.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-JxxkKt9yrx">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/JxxkKt9yrx.html">Oops, I Sampled it Again: Reinterpreting Confidence Intervals in Few-Shot Learning</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Raphael Lafargue · Luke A Smith · Franck VERMET · Matthias Löwe · Ian Reid · Jack Valmadre · Vincent Gripon</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-JxxkKt9yrx"></div>
<a href="paper_pages/JxxkKt9yrx.html">
<img src="http://img.youtube.com/vi/OB95-BLxE0s/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-JxxkKt9yrx" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-JxxkKt9yrx" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-JxxkKt9yrx">
Abstract <i id="caret-JxxkKt9yrx" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-JxxkKt9yrx">
<div class="abstract-display">
<p>The predominant method for computing confidence intervals (CI) in few-shot learning (FSL) is based on sampling the tasks with replacement, i.e. allowing the same samples to appear in multiple tasks. This makes the CI misleading in that it takes into account the randomness of the sampler but not the data itself. To quantify the extent of this problem, we conduct a comparative analysis between CIs computed with and without replacement. These reveal a notable underestimation by the predominant method. This observation calls for a reevaluation of how we interpret confidence intervals and the resulting conclusions in FSL comparative studies. Our research demonstrates that the use of paired tests can partially address this issue. Additionally, we explore methods to further reduce the (size of the) CI by strategically sampling tasks of a specific size. We also introduce a new optimized benchmark, which can be accessed at https://github.com/RafLaf/FSL-benchmark-again</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-A6D3PYSyqJ">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/A6D3PYSyqJ.html">LINOCS: Lookahead Inference of Networked Operators for Continuous Stability</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Noga Mudrik · Eva Yezerets · Yenho Chen · Christopher John Rozell · Adam Shabti Charles</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-A6D3PYSyqJ"></div>
<a href="paper_pages/A6D3PYSyqJ.html">
<img src="http://img.youtube.com/vi/5XVYRHd5wGs/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-A6D3PYSyqJ" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-A6D3PYSyqJ" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-A6D3PYSyqJ">
Abstract <i id="caret-A6D3PYSyqJ" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-A6D3PYSyqJ">
<div class="abstract-display">
<p>Identifying latent interactions within complex systems is key to unlocking deeper insights into their operational dynamics, including how their elements affect each other and contribute to the overall system behavior. For instance, in neuroscience, discovering neuron-to-neuron interactions is essential for understanding brain function; in ecology, recognizing the interactions among populations is key for understanding complex ecosystems. Such systems, often modeled as dynamical systems, typically exhibit noisy high-dimensional and non-stationary temporal behavior that renders their identification challenging. Existing dynamical system identification methods often yield operators that accurately capture short-term behavior but fail to predict long-term trends, suggesting an incomplete capture of the underlying process. Methods that consider extended forecasts (e.g., recurrent neural networks) lack explicit representations of element-wise interactions and require substantial training data, thereby failing to capture interpretable network operators. Here we introduce Lookahead-driven Inference of Networked Operators for Continuous Stability (LINOCS), a robust learning procedure for identifying hidden dynamical interactions in noisy time-series data. LINOCS integrates several multi-step predictions with adaptive weights during training to recover dynamical operators that can yield accurate long-term predictions. We demonstrate LINOCS' ability to recover the ground truth dynamical operators underlying synthetic time-series data for multiple dynamical systems models (including linear, piece-wise linear, time-changing linear systems' decomposition, and regularized linear time-varying systems) as well as its capability to produce meaningful operators with robust reconstructions through various real-world examples.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-xI6cPQObp0">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/xI6cPQObp0.html">Incremental Spatial and Spectral Learning of Neural Operators for Solving Large-Scale PDEs</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Robert Joseph George · Jiawei Zhao · Jean Kossaifi · Zongyi Li · Anima Anandkumar</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-xI6cPQObp0"></div>
<a href="paper_pages/xI6cPQObp0.html">
<img src="https://drive.google.com/thumbnail?id=1rCBmHPxSe_kVAdYX-PntVthA0g6nBYYv" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-xI6cPQObp0" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-xI6cPQObp0" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-xI6cPQObp0">
Abstract <i id="caret-xI6cPQObp0" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-xI6cPQObp0">
<div class="abstract-display">
<p>Fourier Neural Operators (FNO) offer a principled approach to solving challenging partial differential equations (PDE) such as turbulent flows. At the core of FNO is a spectral layer that leverages a discretization-convergent representation in the Fourier domain, and learns
weights over a fixed set of frequencies. However, training FNO presents two significant challenges, particularly in large-scale, high-resolution applications: (i) Computing Fourier transform on high-resolution inputs is computationally intensive but necessary since fine-scale details are needed for solving many PDEs, such as fluid flows, (ii) selecting the relevant set of frequencies in the spectral layers is challenging, and too many modes can lead to overfitting, while too few can lead to underfitting. To address these issues, we introduce
the Incremental Fourier Neural Operator (iFNO), which progressively increases both the number of frequency modes used by the model as well as the resolution of the training data. We empirically show that iFNO reduces total training time while maintaining or improving
generalization performance across various datasets. Our method demonstrates a 38% lower testing error, using 20% fewer frequency modes compared to the existing FNO, while also achieving up to 46% faster training and a 2.8x reduction in model size.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-38P40gJPrI">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/38P40gJPrI.html">Simple Drop-in LoRA Conditioning on Attention Layers Will Improve Your Diffusion Model</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Joo Young Choi · Jaesung R. Park · Inkyu Park · Jaewoong Cho · Albert No · Ernest K. Ryu</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-38P40gJPrI"></div>
<a href="paper_pages/38P40gJPrI.html">
<img src="http://img.youtube.com/vi/z-1LE2aNHak/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-38P40gJPrI" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-38P40gJPrI" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-38P40gJPrI">
Abstract <i id="caret-38P40gJPrI" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-38P40gJPrI">
<div class="abstract-display">
<p>Current state-of-the-art diffusion models employ U-Net architectures containing convolutional and (qkv) self-attention layers. The U-Net processes images while being conditioned on the time embedding input for each sampling step and the class or caption embedding input corresponding to the desired conditional generation. Such conditioning involves scale-and-shift operations to the convolutional layers but does not directly affect the attention layers. While these standard architectural choices are certainly effective, not conditioning the attention layers feels arbitrary and potentially suboptimal. In this work, we show that simply adding LoRA conditioning to the attention layers without changing or tuning the other parts of the U-Net architecture improves the image generation quality. For example, a drop-in addition of LoRA conditioning to EDM diffusion model yields FID scores of 1.91/1.75 for unconditional and class-conditional CIFAR-10 generation, improving upon the baseline of 1.97/1.79.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-30C9AWBW49">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/30C9AWBW49.html">Perceptual Similarity for Measuring Decision-Making Style and Policy Diversity in Games</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Chiu-Chou Lin · Wei-Chen Chiu · I-Chen Wu</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-30C9AWBW49"></div>
<a href="paper_pages/30C9AWBW49.html">
<img src="http://img.youtube.com/vi/eR0Kx6xnH6Y/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-30C9AWBW49" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-30C9AWBW49" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-30C9AWBW49">
Abstract <i id="caret-30C9AWBW49" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-30C9AWBW49">
<div class="abstract-display">
<p>Defining and measuring decision-making styles, also known as playstyles, is crucial in gaming, where these styles reflect a broad spectrum of individuality and diversity. However, finding a universally applicable measure for these styles poses a challenge. Building on $\textit{Playstyle Distance}$, the first unsupervised metric to measure playstyle similarity based on game screens and raw actions by identifying comparable states with discrete representations for computing policy distance, we introduce three enhancements to increase accuracy: multiscale analysis with varied state granularity, a perceptual kernel rooted in psychology, and the utilization of the intersection-over-union method for efficient evaluation. These innovations not only advance measurement precision but also offer insights into human cognition of similarity. Across two racing games and seven Atari games, our techniques significantly improve the precision of zero-shot playstyle classification, achieving an accuracy exceeding 90\% with fewer than 512 observation-action pairs—less than half an episode of these games. Furthermore, our experiments with $\textit{2048}$ and $\textit{Go}$ demonstrate the potential of discrete playstyle measures in puzzle and board games. We also develop an algorithm for assessing decision-making diversity using these measures. Our findings improve the measurement of end-to-end game analysis and the evolution of artificial intelligence for diverse playstyles.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-cD209UgOX7">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/cD209UgOX7.html">Scaling Up Bayesian Neural Networks with Neural Networks</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Zahra Moslemi · Yang Meng · Shiwei Lan · Babak Shahbaba</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-cD209UgOX7"></div>
<a href="paper_pages/cD209UgOX7.html">
<img src="https://drive.google.com/thumbnail?id=10cIAASi5ZN1O5MCm8ojIuTGZhEmygwE4" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-cD209UgOX7" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-cD209UgOX7" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-cD209UgOX7">
Abstract <i id="caret-cD209UgOX7" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-cD209UgOX7">
<div class="abstract-display">
<p>Bayesian Neural Networks (BNNs) offer a principled and natural framework for proper uncertainty quantification in the context of deep learning. They address the typical challenges associated with conventional deep learning methods, such as data insatiability, ad-hoc nature, and susceptibility to overfitting. However, their implementation typically either relies on Markov chain Monte Carlo (MCMC) methods, which are characterized by their computational intensity and inefficiency in a high-dimensional space, or variational inference methods, which tend to underestimate uncertainty. To address this issue, we propose a novel Calibration-Emulation-Sampling (CES) strategy to significantly enhance the computational efficiency of BNN. In this framework, during the initial calibration stage, we collect a small set of samples from the parameter space. These samples serve as training data for the emulator, which approximates the map between parameters and posterior probability. The trained emulator is then used for sampling from the posterior distribution at substantially higher speed compared to the standard BNN. Using simulated and real data, we demonstrate that our proposed method improves computational efficiency of BNN, while maintaining similar performance in terms of prediction accuracy and uncertainty quantification.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-cT8oOJ6Q6F">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/cT8oOJ6Q6F.html">Grid Cell-Inspired Fragmentation and Recall for Efficient Map Building</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Jaedong Hwang · Zhang-Wei Hong · Eric R Chen · Akhilan Boopathy · Pulkit Agrawal · Ila R Fiete</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-cT8oOJ6Q6F"></div>
<a href="paper_pages/cT8oOJ6Q6F.html">
<img src="http://img.youtube.com/vi/lV0clkUssvs/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-cT8oOJ6Q6F" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-cT8oOJ6Q6F" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-cT8oOJ6Q6F">
Abstract <i id="caret-cT8oOJ6Q6F" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-cT8oOJ6Q6F">
<div class="abstract-display">
<p>Animals and robots navigate through environments by building and refining maps of space. These maps enable functions including navigation back to home, planning, search and foraging. Here, we use observations from neuroscience, specifically the observed fragmentation of grid cell map in compartmentalized spaces, to propose and apply the concept of Fragmentation-and-Recall (FARMap) in the mapping of large spaces. Agents solve the mapping problem by building local maps via a surprisal-based clustering of space, which they use to set subgoals for spatial exploration. Agents build and use a local map to predict their observations; high surprisal leads to a "fragmentation event" that truncates the local map. At these events, the recent local map is placed into long-term memory (LTM) and a different local map is initialized. If observations at a fracture point match observations in one of the stored local maps, that map is recalled (and thus reused) from LTM. The fragmentation points induce a natural online clustering of the larger space, forming a set of intrinsic potential subgoals that are stored in LTM as a topological graph. Agents choose their next subgoal from the set of near and far potential subgoals from within the current local map or LTM, respectively. Thus, local maps guide exploration locally, while LTM promotes global exploration. We demonstrate that FARMap replicates the fragmentation points observed in animal studies. We evaluate FARMap on complex procedurally-generated spatial environments and realistic simulations to demonstrate that this mapping strategy much more rapidly covers the environment (number of agent steps and wall clock time) and is more efficient in active memory usage, without loss of performance.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-sIR8xV7hGl">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/sIR8xV7hGl.html">The Kernel Perspective on Dynamic Mode Decomposition</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Efrain Gonzalez · Moad Abudia · Michael Jury · Rushikesh Kamalapurkar · Joel A Rosenfeld</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-sIR8xV7hGl"></div>
<a href="paper_pages/sIR8xV7hGl.html">
<img src="http://img.youtube.com/vi/PLldiDnQu2phuB64ccOYWxeSBZxq0Gg47g/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-sIR8xV7hGl" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-sIR8xV7hGl" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-sIR8xV7hGl">
Abstract <i id="caret-sIR8xV7hGl" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-sIR8xV7hGl">
<div class="abstract-display">
<p>This manuscript takes a critical look at the interactions between Koopman theory and reproducing kernel Hilbert spaces with an eye towards giving a tighter theoretical foundation for
Koopman based dynamic mode decomposition (DMD), a data driven method for modeling
a nonlinear dynamical system from snapshots. In particular, this paper explores the various
necessary conditions imposed on the dynamics when a Koopman operator is bounded or
compact over a reproducing kernel Hilbert space.
Ultimately, it is determined that for many RKHSs, the imposition of compactness or boundedness on a Koopman operator forces the dynamics to be affine.
However, a numerical method is still recovered in more general cases through the consideration of the Koopman operator as a closed and densely defined operator, which requires
a closer examination of the connection between the Koopman operator and a RKHS. By
abandoning the feature representation of RKHSs, the tools of function theory are brought to
bear, and a simpler algorithm is obtained for DMD than what was introduced in Williams
et al (2016). This algorithm is also generalized to utilize vector valued RKHSs.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-dLaazW9zuF">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/dLaazW9zuF.html">Multi-Fidelity Active Learning with GFlowNets</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Alex Hernández-García · Nikita Saxena · Moksh Jain · Cheng-Hao Liu · Yoshua Bengio</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-dLaazW9zuF"></div>
<a href="paper_pages/dLaazW9zuF.html">
<img src="https://www.dailymotion.com/thumbnail/video/k1k8KKYS67DgFCB516w" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-dLaazW9zuF" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-dLaazW9zuF" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-dLaazW9zuF">
Abstract <i id="caret-dLaazW9zuF" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-dLaazW9zuF">
<div class="abstract-display">
<p>In the last decades, the capacity to generate large amounts of data in science and engineering applications has been growing steadily. Meanwhile, machine learning has progressed to become a suitable tool to process and utilise the available data. Nonetheless, many relevant scientific and engineering problems present challenges where current machine learning methods cannot yet efficiently leverage the available data and resources. For example, in scientific discovery, we are often faced with the problem of exploring very large, structured and high-dimensional spaces. Moreover, the high fidelity, black-box objective function is often very expensive to evaluate. Progress in machine learning methods that can efficiently tackle such challenges would help accelerate currently crucial areas such as drug and materials discovery. In this paper, we propose a multi-fidelity active learning algorithm with GFlowNets as a sampler, to efficiently discover diverse, high-scoring candidates where multiple approximations of the black-box function are available at lower fidelity and cost. Our evaluation on molecular discovery tasks shows that multi-fidelity active learning with GFlowNets can discover high-scoring candidates at a fraction of the budget of its single-fidelity counterpart while maintaining diversity, unlike RL-based alternatives. These results open new avenues for multi-fidelity active learning to accelerate scientific discovery and engineering design.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-TAvGZm2Rqb">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/TAvGZm2Rqb.html">Cost-Sensitive Learning to Defer to Multiple Experts with Workload Constraints</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Jean Vieira Alves · Diogo Leitão · Sérgio Jesus · Marco O. P. Sampaio · Javier Liébana · Pedro Saleiro · Mario A. T. Figueiredo · Pedro Bizarro</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-TAvGZm2Rqb"></div>
<a href="paper_pages/TAvGZm2Rqb.html">
<img src="http://img.youtube.com/vi/R51SBj02RuU/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-TAvGZm2Rqb" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-TAvGZm2Rqb" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-TAvGZm2Rqb">
Abstract <i id="caret-TAvGZm2Rqb" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-TAvGZm2Rqb">
<div class="abstract-display">
<p>Learning to defer (L2D) aims to improve human-AI collaboration systems by learning how to defer decisions to humans when they are more likely to be correct than an ML classifier. Existing research in L2D overlooks key real-world aspects that impede its practical adoption, namely: i) neglecting cost-sensitive scenarios, where type I and type II errors have different costs; ii) requiring concurrent human predictions for every instance of the training dataset; and iii) not dealing with human work-capacity constraints. To address these issues, we propose the \textit{deferral under cost and capacity constraints framework} (DeCCaF). DeCCaF is a novel L2D approach, employing supervised learning to model the probability of human error under less restrictive data requirements (only one expert prediction per instance) and using constraint programming to globally minimize the error cost, subject to workload limitations. We test DeCCaF in a series of cost-sensitive fraud detection scenarios with different teams of 9 synthetic fraud analysts, with individual work-capacity constraints. The results demonstrate that our approach performs significantly better than the baselines in a wide array of scenarios, achieving an average $8.4\%$ reduction in the misclassification cost. The code used for the experiments is available at https://github.com/feedzai/deccaf</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-WzHuebRSgQ">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/WzHuebRSgQ.html">A Greedy Hierarchical Approach to Whole-Network Filter-Pruning in CNNs</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Kiran Purohit · Anurag Reddy Parvathgari · Sourangshu Bhattacharya</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-WzHuebRSgQ"></div>
<a href="paper_pages/WzHuebRSgQ.html">
<img src="http://img.youtube.com/vi/egk0ZJb89Ao/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-WzHuebRSgQ" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-WzHuebRSgQ" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-WzHuebRSgQ">
Abstract <i id="caret-WzHuebRSgQ" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-WzHuebRSgQ">
<div class="abstract-display">
<p>Deep convolutional neural networks (CNNs) have achieved impressive performance in many computer vision tasks. However, their large model sizes require heavy computational resources, making pruning redundant filters from existing pre-trained CNNs an essential task in developing efficient models for resource-constrained devices. Whole-network filter pruning algorithms prune varying fractions of filters from each layer, hence providing greater flexibility. State-of-the-art whole-network pruning methods are either computationally expensive due to the need to calculate the loss for each pruned filter using a training dataset, or use various heuristic / learned criteria for determining the pruning fractions for each layer. Hence there is a need for a simple and efficient technique for whole network pruning. This paper proposes
a two-level hierarchical approach for whole-network filter pruning which is efficient and uses the classification loss as the final criterion. The lower-level algorithm (called filter-pruning) uses a sparse-approximation formulation based on linear approximation of filter weights. We explore two algorithms: orthogonal matching pursuit-based greedy selection and a greedy backward pruning approach. The backward pruning algorithm uses a novel closed-form error criterion for efficiently selecting the optimal filter at each stage, thus making the whole algorithm much faster. The higher-level algorithm (called layer-selection) greedily selects the best-pruned layer (pruning using the filter-selection algorithm) using a global pruning criterion. We propose algorithms for two different global-pruning criteria: (1) layerwise-relative error (HBGS), and (2) final classification error (HBGTS). Our suite of algorithms outperforms state-of-the-art pruning methods on ResNet18, ResNet32, ResNet56, VGG16, and ResNext101. Our method reduces the RAM requirement for ResNext101 from 7.6 GB to 1.5 GB and achieves a 94% reduction in FLOPS without losing accuracy on CIFAR-10.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-PLIt3a4yTm">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/PLIt3a4yTm.html">Training-free linear image inverses via flows</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Ashwini Pokle · Matthew J. Muckley · Ricky T. Q. Chen · Brian Karrer</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-PLIt3a4yTm"></div>
<a href="paper_pages/PLIt3a4yTm.html">
<img src="tmlr_logo.jpeg" class="social-img-thumb rounded" alt="thumbnail"/>
</a>
<div class="abstract-section">
<div>
<a id="abstract-link-PLIt3a4yTm" class="abstract-link" data-toggle="collapse"
href="#collapse-event-abstract-PLIt3a4yTm" role="button"
aria-expanded="false" aria-controls="collapse-event-abstract-PLIt3a4yTm">
Abstract <i id="caret-PLIt3a4yTm" class="fas fa-caret-right"></i>
</a>
</div>
</div>
<div class="collapse" id="collapse-event-abstract-PLIt3a4yTm">
<div class="abstract-display">
<p>Solving inverse problems without any training involves using a pretrained generative model and making appropriate modifications to the generation process to avoid finetuning of the generative model. While recent methods have explored the use of diffusion models, they still require the manual tuning of many hyperparameters for different inverse problems. In this work, we propose a training-free method for solving linear inverse problems by using pretrained flow models, leveraging the simplicity and efficiency of Flow Matching models, using theoretically-justified weighting schemes, and thereby significantly reducing the amount of manual tuning. In particular, we draw inspiration from two main sources: adopting prior gradient correction methods to the flow regime, and a solver scheme based on conditional Optimal Transport paths. As pretrained diffusion models are widely accessible, we also show how to practically adapt diffusion models for our method. Empirically, our approach requires no problem-specific tuning across an extensive suite of noisy linear inverse problems on high-dimensional datasets, ImageNet-64/128 and AFHQ-256, and we observe that our flow-based method for solving inverse problems improves upon closely-related diffusion-based methods in most settings.</p>
</div>
</div>
</div>
<div class="displaycards touchup-date" id="event-DimPeeCxKO">
<div style="width:80%;margin:auto;">
<a class="small-title" href="paper_pages/DimPeeCxKO.html">Simple and Scalable Strategies to Continually Pre-train Large Language Models</a>
</div>
<div class="type_display_name_minus_type"></div>
<div class="author-str">Adam Ibrahim · Benjamin Thérien · Kshitij Gupta · Mats Leon Richter · Quentin Gregory Anthony · Eugene Belilovsky · Timothée Lesort · Irina Rish</div>
<div class="author-str higher"></div>
<div class="text-muted touchup-date-div" id="touchup-date-event-DimPeeCxKO"></div>
<a href="paper_pages/DimPeeCxKO.html">
<img src="http://img.youtube.com/vi/y4sUn3sYWFc/0.jpg" class="social-img-thumb rounded" alt="thumbnail"/>