How mathangi18/lab-2 works
CM2003 Lab 2 is a hands-on neural-network assignment, delivered through notebooks and a NumPy support package. Learners load CIFAR-10 or PneumoniaMNIST, implement and train fully connected or convolutional models, then evaluate results; the PyTorch notebook additionally covers medical-data metrics and training practices. The graph includes the documented optional batch-normalization and dropout exercises. Notebook internals were not sampled, so their links to package components are limited to relationships the README or source index establishes.
Main entry points: cnn.py Convolutional model · layer_utils.py Composite layer helpers · fast_layers.py Fast convolution and pooling · solver.py Training solver
Components 20
Lab notebooks
- Fully connected lab
FullyConnectedNets.ipynb - Batch normalization lab
BatchNormalization.ipynb - Dropout lab
Dropout.ipynb - Convolutional lab
ConvolutionalNetworks.ipynb - PyTorch medical lab
PyTorch.ipynb
NumPy modeling
- Fully connected models
cm2003/classifiers/fc_net.py - Convolutional model
cm2003/classifiers/cnn.py - Layer operations
cm2003/layers.py - Composite layer helpers
cm2003/layer_utils.py - Fast convolution and pooling
cm2003/fast_layers.py - Im2col transforms
cm2003/im2col.py - Training solver
cm2003/solver.py - Update rules
cm2003/optim.py
Data and utilities
- CIFAR data helpers
cm2003/data_utils.py - CIFAR-10 and PneumoniaMNIST
cm2003/datasets - Dataset downloader
download_data.py - Gradient checking
cm2003/gradient_check.py - Image utilities
cm2003/image_utils.py - Visualization utilities
cm2003/vis_utils.py
Other components
- Learner
Connections 20
- Learner → Fully connected lab: works through
README.md - Learner → Batch normalization lab: optionally implements
README.md - Learner → Dropout lab: explores
README.md - Learner → Convolutional lab: implements
README.md - Learner → PyTorch medical lab: trains and evaluates
README.md - Dataset downloader → CIFAR-10 and PneumoniaMNIST: downloads or copies
download_data.py - Fully connected lab → CIFAR-10 and PneumoniaMNIST: loads CIFAR-10
README.md - PyTorch medical lab → CIFAR-10 and PneumoniaMNIST: loads PneumoniaMNIST
README.md - Fully connected models → Layer operations: uses
cm2003/classifiers/fc_net.py - Fully connected models → Composite layer helpers: uses
cm2003/classifiers/fc_net.py - Convolutional model → Layer operations: uses
cm2003/classifiers/cnn.py - Convolutional model → Fast convolution and pooling: uses
cm2003/classifiers/cnn.py - Convolutional model → Composite layer helpers: uses
cm2003/classifiers/cnn.py - Composite layer helpers → Layer operations: calls
cm2003/layer_utils.py - Composite layer helpers → Fast convolution and pooling: calls
cm2003/layer_utils.py - Fast convolution and pooling → Im2col transforms: uses transforms
cm2003/fast_layers.py - Training solver → Update rules: dispatches updates
cm2003/solver.py - Image utilities → Fast convolution and pooling: uses convolution
cm2003/image_utils.py - Training solver → Fully connected models: trains model
- Training solver → Convolutional model: trains model