This repository contains the projects done during my UCL MSc Machine Learning course
Folder content:
СОМР0078CW1 - First coursework of the module СОМР0078 Supervised Learning. In this coursework I implement polynomial regression, kernel ridge regression, k-nearest-neighbours and analyze them from a theoretical point of view.
СОМР0078CW2 - Second coursework of the module СОМР0078 Supervised Learning. In this coursework I implement a kernel perceptron from scratch using one-vs-rest and one-vs-all method for classification, I analyze spectral clustering and finally I analyze the sample complexity of the k-nn, perceptron, regression and winnow algorithms.
СОМР0163CW1 - First coursework of the module СОМР0163 Blockchain Technologies. In this coursework I wrote a smart contract in DAML "DAMLCW.pdf" (an open-source smart contract language) as well as a smart contract in Solidity "SmartContract.sol", "SolCW.pdf".
СОМР0163CW2 - First coursework of the module СОМР0163 Blockchain Technologies. In this group coursework we develop a smart contract for options and analyze them in the context of Decentralized Exchanges.
СОМР0090CW1 - First coursework of the module СОМР0090 Deep Learning. In this coursework I implement a stochastic gradient descent algorithm, a DenseNet and implement an ablation study.
COMP0081CW1 - First coursework of the module COMP0081. Here I implemented a classification and regression tree, adaboost and random forest from scratch.
COMP0080CW - Coursework for the module COMP0080 Graphical Models. The principles of Inference and Learning were applied in the problems of double earthquakes, meeting scheduling and weather stations’ data sampling. The encoding and decoding of LDPC-codes, using Loopy Belief Propagation for Binnary Symmetric Channel, was explored. Exact inference, the Mean Field Approximation and Gibbs Sampling were applied, to find the joint probability distribution of the Ising model on a 10 by 10 lattice.
RandomForests.ipynb - First Coursework of COMP0081 Applied Machine Learning.
СОМР0084CW1 - First coursework of the module СОМР0084 Information Retrieval and Data Mining. Here I implemented an inverted index. Using the index, I ranked groups of passages to certain queries using the Cosine Similarity and BM25 Metric. I also use Laplace Smoothing, Lidstone correction and Dirchlet Smoothing.
СОМР0084CW2 - Second coursework of the module СОМР0084 Information Retrieval and Data Mining. Here I implemented various algorithms which evaluated passage-ranking algorithms. Some of the algorithms include Logistic regression, Neural Network and LambdaMart.
COMP0089 - Coursework for the module COMP0089 Reinforcement Learning. Here I implement various RL algorithms such as bandits, Temporal Difference Learning and Monte Carlo.