Mar 2025
Experiments with Machine Learning
Top: output from a CNN trained to classify images. Middle: output from an RNN trained on Shakespeare’s sonnets. Bottom left: output from a DNN perceptron trained to identify hand-written numbers. Bottom right: screenshot of a grid world RL environment.
Machine learning (ML) is transforming the ways that humans work together with computers. Instead of following rigid rules, ML identifies patterns, makes predictions, and improves its performance over time.
The applications of machine learning are very exciting to me, so I explored applying ML methods in the context of some standard problems:
A deep neural network (DNN) perceptron for identifying hand-written numbers from the MNIST dataset.
A recurrent neural network (RNN) for generating text, based off learning from Shakespeare’s plays and sonnets.
A convolutional neural network (CNN) for classifying images from the CIFAR-10 dataset.
A tabular Q-learning implementation of reinforcement learning (RL) to solving a grid world environment.