CS550 Machine Learning

The assignment and project implementations for CS550 Machine Learning course, Bilkent University.

Coverage of the repository

HW1

Decision Trees

• Used sklearn to train an decision tree classifier to classify thyroid disease patients.
• Implemented a decision tree classifier to build a decision tree with preprunning.
• Extended the implementation for a Cost-Sensitive decision tree.
• Implemented in Python (Jupyter Notebook)

HW2

Linear Regression and Neural Networks

• Trained a neural network and a linear regressor and optimize to learn the correlation between the provided input output sets. Implemented a general purpose neural network framework under backend, however, this implementation is not completely accurate. I wouldn't recommend anyone to clone this repo and use that backend. The future work consists of debugging.
• Backend is implemented in Python, and the experiments are done in Python using Jupiter Notebooks

HW3

Clustering

• Implemented k-Means clustering algorithm to cluster the image pixels in an image.
• Implemented Hierarchical Agglomerative Clustering (HAC) to cluster the pixels. Since this is a expensive algorithm in terms of memory and time, we have initially clustered the pixels with k-Means, then used HAC to cluster the rest.
• Implemented in Python (Jupyter Notebook)

Project:

Image Domain Adaptation using Cyclic Generative Adversarial Networks

• Used Cycle GANs to transform the images from the photograph to cartoon domain.
• Custom framed and cropped face cartoon dataset is used.
• Experimented on Vanilla GAN, LSGAN and Wasserstein GAN architectures.
• Used this repo to train the networks.
• Frechet Inception Distance (FID) implemented to measure performance.