BrainNet

This GitHub repository contains coding files for my MSc Project, entitled
Approximation of Arbitrary Neuronal Network with Deep Learning
Submitted as partial requirements of the MSc in Machine Learning in University College London.

All files are in the masters branch:

• approx_bnn.py contains 4 classes of approximate biological neural networks (ABNNs);
• gen_data.py contains code for generating synthetic spike trains and input patterns to feed into ABNNs;
• models.py contains DNN classes: MLP and RNN;
• train.py contains main training loop;
• bvc.py contains classes for the environment, roaming agent, the boundary vector cell and place cell network;
• utils.py contains a range of auxiliary functions, such as plotting;
  Folders:
• approx_bnn_params stores parameters for ABNNs, including individual transfer functions for each neuron;
• dnn_params stores parameters for trained DNNs;
• data stores input-output pairs generated from synthetic spike trains, ABNN and BVC network;
• temp stores temporary files and training results for plotting;
• figures stores saved figures
  Notebooks:
• abnn_pattern_analysis contains experiments on analysing synthetic neuronal inputs and outputs;
• abnn_experiments contains all experiments in the first part (ABNN);
• bvc_experiments contains all experiments in the second part (BVC model).
  To recreate the results, run
  git clone https://github.com/cngzlsh/BrainNet.git --branch master
  And use the seed (1234) embedded at the beginning of each file.
  The codes are tested on a machine with the following specifications:
• Intel Core i7-11370H CPU (4-core, 8-threads), 40 GiB RAM, Nvidia RTX 3070 Laptop GPU (5,120-core, 8 GiB DRAM), 512 GiB SSD on Pop!_OS 22.04 LTS (Gnome version 42.4).