Unsupervised Feature Learning and Transfer Analysis in Computer Vision: Conducted a comprehensive evaluation of unsupervised feature learning methods using CIFAR-10 and DomainNet datasets, followed by an investigation into transfer learning in computer vision with a focus on adapting to new domains.
RNN-Based Sentiment Classification and Generalization Analysis: Developed and assessed RNN models for sentiment classification, examining architectural variations and Word2Vec embeddings. Analyzed model generalization by retraining on a modified dataset and evaluating performance on IMDB data.
Deep Learning for Image Classification and Feature Extraction: Trained ResNet, DenseNet, and VGG16 neural networks on dog and cat datasets, optimizing accuracy by experimenting with different percentages of frozen layers. Leveraged the networks as feature extractors for data classification using random forests.
MNIST Neural Network Classification and Adaptation: Developed and fine-tuned a neural network for MNIST classification, including layer and optimizer experimentation. Evaluated key metrics and assessed network adaptability to new classes, with an outlook on potential outlier detection strategies.
Step by step ROS2 course consisting of guided tutorials and exercises based on official ROS2 and Gazebo documentations.
Reinforcement Learning Projects
Spring 2023
Dynamic Strategies in Frozen Lake: Executed value and policy surveys in GymOpenAI’s Frozen Lake, exploring factors like discount variations and non-deterministic problem-solving, visualizing outcomes with Matplotlib.
Algorithmic Exploration in Taxi World: Implemented Q-Learning, Monte Carlo, and Sarsa algorithms in Gym’s Taxi environment. Visualized diverse gamma scenarios, charted cumulative rewards, and discerned differences between Q-Learning and Sarsa approaches.