Blog
This post details initial release of website - why it exists and how it's made.
Archive: First steps with Convolutional Networks
Implementing convolutional neural networks forwards and backwards passes with Numpy, and convolution weights visualization/interpretation.
Archive: Dropout and batch normalization
Implementing dropout and batch norm to make training models easier and improve their evaluation performance.
Archive: More with fully-connected nets
Expanding the PyTorch-inspired neural net programming interface to accept arbitrary model shapes and additional training optimization schemes.
Archive: Feature extraction and fully-connected neural networks
First steps with feature extraction, followed by making a reusable neural network interface which loosely mirrors PyTorch's design.
Archive: Two layer neural net
How I made my first neural network to perform image classification on CIFAR-10.
Archive: Softmax classifier
How I derived and developed a linear softmax classifier on the CIFAR-10 dataset.
Archive: Implementing KNN and Support Vector Machine classifiers
The first post in the deep learning series, this marks the beginning of my journey to get up to speed with the latest computer vision literature.