- CA1: Genetics Algorithm
- CA2: Hidden Markov Model
- CA3: Clustering
- CA4: Machine Learning
- CA5: Convolutional Neural Network
- CA6: Reinforcement Learning/snake_code
In this project, we aim to implement one of the optimization problems using genetic algorithms. The goal of an optimization problem is to find the best solution from a large set of possible solutions, where the "best" solution is defined by a set of criteria or objectives. Solving optimization problems can lead to significant benefits such as increased efficiency, cost reduction, improved performance, and more.
Hidden Markov Model
The main objective of this project was to develop a speech recognition system for numbers using Hidden Markov Models (HMM). I was given an audio dataset, conducted preprocessing steps, extracted relevant features, implemented the HMM model, and performed a thorough analysis of the results to evaluate the system's performance.
In this project, we aim to analyze images of various flowers using clustering algorithms. Our goal is to categorize them into different groups based on the available data, so that after applying the clustering algorithm, the flowers are placed in the correct clusters as accurately as possible.
The project involves three main sections: first, exploring and understanding the data, including its distribution and statistical details. Second, the critical step of data preprocessing, where real-world data is transformed for effective use in machine learning models. Finally, different machine learning models are created and evaluated.
In this project, I worked with textual data, starting with text preprocessing. I then utilized the word2vec model to prepare sentences for input into a neural network. I implemented a convolutional neural network using the PyTorch library for sentence classification and observed the effect of text window size on the network's learning. Additionally, I examined the impact of regularization methods on the training process.
The challenge involves using Q-learning to train an AI agent to play effectively. Given the extensive state space, efficient feature extraction and representation methods are crucial. The approach should involve defining features that accurately describe the game environment, allowing for effective training of the model.