-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
275 lines (234 loc) · 17.9 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
<!DOCTYPE html>
<html lang="en">
<head>
<!-- Basic Page Needs
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
<meta charset="utf-8">
<title>Antoine Dedieu</title>
<meta name="description" content="">
<meta name="author" content="">
<!-- Mobile Specific Metas
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
<meta name="viewport" content="width=device-width, initial-scale=1">
<!-- FONT
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
<link href="//fonts.googleapis.com/css?family=Raleway:400,300,600" rel="stylesheet" type="text/css">
<!-- CSS
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
<link rel="stylesheet" href="css/normalize.css">
<link rel="stylesheet" href="css/skeleton.css">
<!-- Favicon
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
<link rel="icon" type="image/png" href="images/deepmind.png">
<script type="text/javascript">
(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){
(i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),
m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m)
})(window,document,'script','https://www.google-analytics.com/analytics.js','ga');
ga('create', 'UA-1592615-30', 'auto', 'webwwwall');
ga('webwwwall.send', 'pageview');
</script>
</head>
<body>
<!-- Primary Page Layout
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
<div class="container">
<div class="row">
<div class="eight columns" style="margin-top: 5%">
<h1>Antoine Dedieu</h1>
<p> <h5>
Senior Research Scientist at <a class="hidelink" href="https://www.deepmind.com">DeepMind</a> <br>
<a class="hidelink" href="http://www.mit.edu">MIT</a> - <a class="hidelink" href="https://www.polytechnique.edu">Ecole Polytechnique</a>
</h5> </p>
<p> <a class="hidelink" href="https://scholar.google.com/citations?user=Hgoc3FUAAAAJ">Scholar</a> -
<a class="hidelink" href="Antoine_Dedieu.pdf">Resume</a> -
<a class="hidelink" href="https://www.linkedin.com/in/antoine-dedieu-9529a4a9/">LinkedIn</a> -
<a class="hidelink" href="mailto:antoine@deepmind.com">Contact</a></p>
</div>
<div class="four columns" style="margin-top: 5%">
<img src="images/photo2020.jpg" alt="headshot" width="100%">
</div>
</div>
<div class="row">
<div class="twelve columns" style="margin-top: 0%">
<h3>About me</h3>
<ul>
<p align="justify">
I am a Senior Research Scientist at <a class="hidelink" href="https://deepmind.google/">Google DeepMind</a>.
I am currently interested in building agents that can efficiently learn new tasks in new environments.
I believe that such an agent must learn a latent world model, which pairs
(a) a representation model mapping environment observations to a rich compact latent space, with
(b) a generative world model describing the representation dynamics.
I also believe that LLMs / VLMs can be leveraged as rich priors for solving these new tasks.
My research then lies at the intersection of representation learning, model-based reinforcement learning and in-context learning.
<br>
<br>
Prior to joining <a class="hidelink" href="https://deepmind.google/">Google DeepMind</a>,
I worked 3.5 years as a Research Scientist then as a Senior Research Scientist at
<a class="hidelink" href="https://www.vicarious.com">Vicarious</a>
(<a class="hidelink" href="https://intrinsic.ai/blog/posts/mission-momentum/">acquired</a> by <a class="hidelink" href="https://abc.xyz">Alphabet</a>)
where I was building an AI layer for robots.
My research there was focus on building novel generative probabilistic graphical models (PGMs) to solve challenging (a) object-centric vision problems,
and (b) navigation problems, and on deriving new methods for learning and inference in complex PGMs.
<br>
<br>
Prior to joining Vicarious, I graduated from MIT <a class="hidelink" href="http://web.mit.edu/orc/www/">
Operations Research Center</a> Master of Science, where I was advised by Prof. <a class="hidelink" href="http://www.mit.edu/~rahulmaz/">
Rahul Mazumder</a>. My reseach was focused on building new algorithms to compute interpretable estimators, and studying their statistical performance.
Before MIT, I earned an MS in Applied Mathematics from <a class="hidelink" href="https://www.polytechnique.edu">
Ecole Polytechnique</a>.
<!--During that time, I worked as Machine Learning research intern in the
Equity Derivative Structuring team of <a class="hidelink" href="https://cib.societegenerale.com">Societe Generale</a>
and gained two international experiences at <a class="hidelink" href="https://www.option.cl">Option</a>, Santiago (Chile)
and at <a class="hidelink" href="http://speit.sjtu.edu.cn/indexen.html">Shanghai ParisTech Jiao Tong</a> (China).-->
<br>
<br>
I am a movie-fan, a sports addict, and a world-traveller. As an amateur photograph, I share my journeys around on
my <a class="hidelink" href="https://worldroadjourney.wordpress.com">website</a>.
</ul>
<hr>
<h3>Conference articles</h3>
<ul>
<li>
<strong><a class="hidelink" href="https://arxiv.org/pdf/2409.18330">Schema-learning and rebinding as mechanisms of in-context learning and emergence.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2409.18330"> [Preprint]</a>
<a class="hidelink" href="https://github.com/google-deepmind/dmc_vision_benchmark"> [Code]</a>
<br><i>Neurips 2024, to appear.</a></i>
Antoine Dedieu*, Joseph Ortiz*, Wolfgang Lehrach, Swaroop Guntupalli, Carter Wendelken, Ahmad Humayun, Sivaramakrishnan Swaminathan, Guangyao Zhou, Miguel Lázaro-Gredilla, Kevin Murphy.
</li>
<li>
<strong><a class="hidelink" href="https://openreview.net/pdf?id=JUa5XNXuoT">Learning Cognitive Maps from Transformer Representations for Efficient Planning in Partially Observed Environments.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2401.05946"> [Preprint]</a>
<br><i><a class="hidelink" href="https://proceedings.mlr.press/v235/dedieu24a.html">ICML 2023.</a></i>
Antoine Dedieu, Wolfgang Lehrach, Guangyao Zhou, Dileep George, Miguel Lázaro-Gredilla.
</li>
<li>
<strong><a class="hidelink" href="https://proceedings.neurips.cc/paper_files/paper/2023/file/5bc3356e0fa1753fff7e8d6628e71b22-Paper-Conference.pdf">Schema-learning and rebinding as mechanisms of in-context learning and emergence.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2307.01201"> [Preprint]</a>
<br><i><a class="hidelink" href="https://proceedings.neurips.cc/paper_files/paper/2023/hash/5bc3356e0fa1753fff7e8d6628e71b22-Abstract-Conference.html">Neurips 2023, Spotlight.</a></i>
Sivaramakrishnan Swaminathan, Antoine Dedieu, Rajkumar Vasudeva Raju, Murray Shanahan, Miguel Lázaro-Gredilla, Dileep George.
</li>
<li>
<strong><a class="hidelink" href="http://proceedings.mlr.press/v202/dedieu23a/dedieu23a.pdf">Learning noisy-OR Bayesian Networks with Max-Product Belief Propagation.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2302.00099"> [Preprint]</a>
<a class="hidelink" href="https://github.com/google-deepmind/max_product_noisy_or"> [Code]</a>
<br><i><a class="hidelink" href="https://icml.cc/virtual/2023/poster/25126">ICML 2023.</a></i>
Antoine Dedieu, Guangyao Zhou, Dileep George, Miguel Lázaro-Gredilla.
</li>
<li>
<strong><a class="hidelink" href="https://arxiv.org/pdf/2112.03371.pdf">Graphical Models with Attention for Context-Specific Independence and an Application to Perceptual Grouping.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2112.03371"> [Preprint]</a>
<a class="hidelink" href="https://github.com/vicariousinc/mam/"> [Code]</a>
<br>Guangyao Zhou, Wolfgang Lehrach, Antoine Dedieu, Miguel Lázaro-Gredilla, Dileep George.
</li>
<li>
<strong><a class="hidelink" href="https://proceedings.neurips.cc/paper/2021/file/07b1c04a30f798b5506c1ec5acfb9031-Paper.pdf">Perturb-and-max-product: Sampling and learning in discrete energy-based models.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2111.02458"> [Preprint]</a>
<a class="hidelink" href="https://github.com/vicariousinc/perturb_and_max_product"> [Code]</a>
<br><i><a class="hidelink" href="https://proceedings.neurips.cc/paper/2021/hash/07b1c04a30f798b5506c1ec5acfb9031-Abstract.html"> Neurips 2021.</a></i>
Miguel Lázaro-Gredilla, Antoine Dedieu, Dileep George.
</li>
<li>
<strong><a class="hidelink" href="https://ojs.aaai.org/index.php/AAAI/article/view/16884">Sample-Efficient L0-L2 Constrained Structure Learning of Sparse Ising Models.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2012.01744"> [Preprint]</a>
<a class="hidelink" href="https://github.com/antoine-dedieu/structure_learning_sparse_ising_models"> [Code]</a>
<br><i><a class="hidelink" href="https://ojs.aaai.org/index.php/AAAI/article/view/16884">AAAI 2021.</a></i>
Antoine Dedieu, Miguel Lázaro-Gredilla, Dileep George.
</li>
<li>
<strong><a class="hidelink" href="https://ojs.aaai.org/index.php/AAAI/article/view/17004">Query Training: Learning a Worse Model to Infer Better Marginals in Undirected Graphical Models with Hidden Variables.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2006.06803"> [Preprint]</a>
<a class="hidelink" href="https://github.com/vicariousinc/query_training"> [Code]</a>
<br><i> <a class="hidelink" href="https://ojs.aaai.org/index.php/AAAI/article/view/17004">AAAI 2021.</a></i>
Miguel Lázaro-Gredilla, Wolfgang Lehrach, Nishad Gothoskar, Guangyao Zhou, Antoine Dedieu, Dileep George.
</li>
<li>
<strong><a class="hidelink" href="https://arxiv.org/pdf/1910.08880.pdf">Improved error rates for sparse (group) learning with Lipschitz loss functions.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/1910.08880"> [Preprint]</a>
<br>Antoine Dedieu.
</li>
<li>
<strong><a class="hidelink" href="https://arxiv.org/pdf/1912.11398.pdf">An error bound for Lasso and Group Lasso in high dimensions.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/1912.11398"> [Preprint]</a>
<br>Antoine Dedieu.
</li>
<li>
<strong><a class="hidelink" href="https://arxiv.org/pdf/1905.00507.pdf">Learning higher-order sequential structure with cloned HMMs.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/1905.00507"> [Preprint]</a>
<br>Antoine Dedieu, Nishad Gothoskar, Scott Swingle, Wolfgang Lehrach, Miguel Lázaro-Gredilla, Dileep George.
</li>
<li>
<strong><a class="hidelink" href="http://proceedings.mlr.press/v89/dedieu19a/dedieu19a.pdf">Error bounds for sparse classifiers in high-dimensions.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/1810.03081"> [Preprint]</a>
<br><i><a class="hidelink" href="http://proceedings.mlr.press/v89/dedieu19a.html"> AiStats 2019.</a></i>
Antoine Dedieu.
</li>
<li>
<strong><a class="hidelink" href="https://dl.acm.org/doi/10.1145/3269206.3271700">Hierarchical Modeling and Shrinkage for User Session Length Prediction in Media Streaming.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/1803.01440"> [Preprint]</a>
<a class="hidelink" href="https://github.com/pvahabi/MIT_session_length_prediction"> [Code]</a>
<br><i><a class="hidelink" href="https://dl.acm.org/doi/10.1145/3269206.3271700"> CIKM 2018.</a></i>
Antoine Dedieu, Rahul Mazumder, Zhen Zhu, Hossein Vahabi.
</li>
</ul>
<h3>Journal articles</h3>
<ul>
<li>
<strong><a class="hidelink" href="https://arxiv.org/pdf/2202.04110.pdf">PGMax: Factor Graphs for Discrete Probabilistic Graphical Models and Loopy Belief Propagation in JAX.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2202.04110"> [Preprint]</a>
<a class="hidelink" href="https://github.com/google-deepmind/PGMax"> [Code]</a>
<br><i>Journal of Machine Learning Research (accepted with minor revisions), 2023.</i> Guangyao Zhou, Antoine Dedieu, Nishanth Kumar, Miguel Lázaro-Gredilla, Shrinu Kushagra, Dileep George.
</li>
<li>
<strong><a class="hidelink" href="https://www.biorxiv.org/content/10.1101/2020.12.31.424926v1.full">Learning attention-controllable border-ownership for objectness inference and binding.</a></strong>
<a class="hidelink" href="https://www.biorxiv.org/content/10.1101/2020.12.31.424926v1"> [Preprint]</a>
<br>Antoine Dedieu, Rajeev V. Rikhye, Miguel Lázaro-Gredilla, Dileep George.
</li>
<li>
<strong><a class="hidelink" href="https://www.biorxiv.org/content/10.1101/2020.09.09.290601v1.full">A detailed mathematical theory of thalamic and cortical microcircuits based on inference in a generative vision model.</a></strong>
<a class="hidelink" href="https://www.biorxiv.org/content/10.1101/2020.09.09.290601v1"> [Preprint]</a>
<br>Dileep George, Miguel Lázaro-Gredilla, Wolfgang Lehrach, Antoine Dedieu, Guangyao Zhou.
</li>
<li>
<strong><a class="hidelink" href="https://www.jmlr.org/papers/volume23/19-104/19-104.pdf">Solving L1-regularized SVMs and related linear programs: Revisiting the effectiveness of Column and Constraint Generation.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/1901.01585"> [Preprint]</a>
<a class="hidelink" href="https://github.com/antoine-dedieu/cutting_planes_l1_SVM_and_cousins"> [Code]</a>
<br><i><a class="hidelink" href="https://jmlr.org/papers/v23/19-104.html">Journal of Machine Learning Research, 2022.</a></i>
Antoine Dedieu, Rahul Mazumder, Haoyue Wang
</li>
<li>
<strong><a class="hidelink" href="https://arxiv.org/pdf/1708.03288.pdf">Subset Selection with Shrinkage: Sparse Linear Modeling when the SNR is low.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/1708.03288"> [Preprint]</a>
<a class="hidelink" href="https://github.com/antoine-dedieu/subset_selection_with_shrinkage"> [Code]</a>
<br><i><a class="hidelink" href="https://pubsonline.informs.org/doi/abs/10.1287/opre.2022.2276">Operations Research, 2022.</a></i>
Rahul Mazumder, Peter Radchenko, Antoine Dedieu.
</li>
<li>
<strong><a class="hidelink" href="https://jmlr.org/papers/volume22/19-1049/19-1049.pdf">Learning Sparse Classifiers: Continuous and Mixed Integer Optimization Perspectives.</a></strong>
<a class="hidelink" href="https://arxiv.org/abs/2001.06471"> [Preprint]</a>
<a class="hidelink" href="https://github.com/hazimehh/L0Learn"> [Code]</a>
<br><i><a class="hidelink" href="https://jmlr.org/papers/v22/19-1049.html">Journal of Machine Learning Research, 2021.</a></i>
Antoine Dedieu, Hussein Hazimeh, Rahul Mazumder.
</li>
<li>
<strong><a class="hidelink" href="https://www.nature.com/articles/s41467-021-22559-5">Clone-structured graph representations enable flexible learning and vicarious evaluation of cognitive maps.</a></strong>
<a class="hidelink" href="https://www.biorxiv.org/content/10.1101/864421v4"> [Preprint]</a>
<br><i><a class="hidelink" href="https://www.nature.com/articles/s41467-021-22559-5"> Nature Communications, 2021.</a></i>
Dileep George, Rajeev V Rikhye, Nishad Gothoskar, J Swaroop Guntupalli, Antoine Dedieu, Miguel Lázaro-Gredilla.
</li>
</ul>
<h3>Thesis</h3>
<ul>
<li>
<strong><a class="hidelink" href="https://dspace.mit.edu/handle/1721.1/119354">Sparse learning: statistical and optimization perspectives.</a></strong>
<br><i><a class="hidelink" href="https://dspace.mit.edu/handle/1721.1/119354">MIT Libraries.</a></i>
Antoine Dedieu.
</li>
</ul>
</div>
</div>
</div>
<!-- End Document
–––––––––––––––––––––––––––––––––––––––––––––––––– -->
</body>
</html>