Hafez Ghaemi, Erfan Mirzaei, Mahbod Nouri, Saeed Reza Kheradpisheh

DOI: https://doi.org/10.1007/978-3-031-25891-6_42

arXiv: https://arxiv.org/abs/2109.05539

To install bindsnet you should only use the following command:

!pip install -q git+https://github.com/bindsnet/bindsnet

This will downgrade some of your installed packages including PyTorch.

In this part, we trained our hidden layer to extract features from the MNIST images. We use these features as pre-trained weights in the classification task.

After transfering the weights from the pretrained features from the previous section, we train the network to classify the MNIST dataset images. You can also run "main.ipynb" for this experiment.

In this experiment, we present the network with images belonging to one arbitrary class of the MNIST dataset as the neutral stimuli, and give reward such that it becomes conditioned to the desired response during each task. The purpose of this experiment is to show the effectiveness of our rewarding mechanism.