Analog Printed Spiking Neuromorphic Circuit

This github repository is for the paper at DATE'24 - Analog Printed Spiking Neuromorphic Circuit

cite as

Analog Printed Spiking Neuromorphic Circuit
Pal, P.; Zhao, H.; Shatta, M,; Hefenbrock, M.; Mamaghani, S. B.; Nassif, S.; Beigl, M.; Tahoori, M. B.
2024 Design, Automation & Test in Europe Conference & Exhibition (DATE), IEEE, 2024

Usage of the code:

0. Modeling of the printed spiking g

In the folder ./simulation/ locate the data from SPICE simulation based on printed Processing Design Kit (pPDK). Different temporal input signals $x(t)$ are simulated and yielded the corresponding output signal $y(t)$. This part aims to build a machine learning based surroagate model of the printed spiking generator (pSG) with

$$y(t) = \rm{pSG}(x(t)).$$

Simply run the jupyter notebooks one by one

1_read_cascade.ipynb
...
5_visualization.ipynb

1. Training of printed neural networks

After obtaining the machine learning based model of pSG, the whole circuit (including resistor crossbar for weighted-sum and pSG for nonlinearity) can be trained through

$ sh run_pSNN.sh

Alternatively, the experiments can be conducted by running command lines in exp_pSNN.sh separately, e.g.,

$ python3 exp_pSNN.py --DATASET 00 --SEED 0 --projectname pSNN
$ python3 exp_pSNN.py --DATASET 00 --SEED 1 --projectname pSNN
$ python3 exp_pSNN.py --DATASET 00 --SEED 2 --projectname pSNN
...

Analogous for baselines, the circuit can be trained through

$ sh run_SNN.sh

and

$ sh run_pNN.sh

2. After training printed neural networks, the trained networks are in ./pSNN/model/, the log files for training can be found in ./pSNN/log/. If there is still files in ./pSNN/temp/, you should run the corresponding command line to train the networks further. Note that, each training is limited to 48 hours, you can change this time limitation in configuration.py

3. Evaluation can be done by running the Evaluation_pSNN.ipynb in ./pSNN/ folder with