Kuldeep Gautam

• Logistic Regression Numpy based implementation of logistic regression with log loss as the loss function. It includes utility for fit, predict and score functions, similar to scikit-learn API.
• Decision Trees Implementation of entropy and gini-index based decision tree classifier. It also includes implementation of K-Fold cross validation from scratch.
• Random Forests Implementation for entropy and gini-index based random forest classifier, making use of the decision tree module built in the above project. It also provides utility functions for fitting, predictions, out of the bag score and performs at par with scikit-learn RFC API in accuracy. Sensitivity to max_features parameter and OOB error has also been explored.
• Feed Forward Neural Networks Tensorflow based implementation of FF-NN. Adam optimizer has been used to perform the backward pass only using TF API. Results in an accuracy of 96.6% on MNIST dataset.
• Density-Based Spatial Clustering of Applications with Noise (DBSCAN) Implementation of DBSCAN method of clustering. Used euclidean distance for getting points in epsilon neighborhood and separates the points into core, border and noise points. Only core points and border points are used to define boundary lines for clusters. Performs at par with scikit learn implementaton.