Predicted tags for a StackOverflow dataset. Given Questions and Body, we need to predict the tags for each Questions presented. Predicting by using TF-IDF and Logistic Regression with OneVsRestClassifier Predicting with LSTM and GRU models
• Deep Q-Learning on Banana Navigation Project Implemented deep Q-learning model with replay buffer and experience replay to train the agent • Deep Deterministic Policy Gradient (DDPG) for Reacher environment Implemented the DDPG in continuous action space that use actor-critic and model-free algorithms • Multi Agent DDPG for Tennis environment Implemented the MADDPG for Tennis environment Used the first agent’s experience to train the second agent.
2019-8 - 2019-11
Histopathologic Image Classification
• Master student’s theses project • Finding the most suitable optimizer and activation function to improve the classification accuracy • Use convolutional neural networks and transfer learning from various models available
Deep Learning with Amazon SageMaker Implemented Sentiment analysis web applications by using LSTM and deployed in Amazon SageMaker Face Generation Generated new faces from celebrity’s dataset Used Deep Convolutional Generative Adversarial Network (DCGAN) TV Script Generation Generated TV scripts by using Recurrent Neural Networks Used Long-Short Term Memory model Dog Breed Classification Used CNN and transfer learning as classifier Flower Classification Classified images based on pretrained models Applied the convolutional neural networks and deep learning frameworks.
2018-11 - 2019-5
Graph based Transform for Pathological Image Compression
A part of research internship project in University of Warwick, UK The compression system used is HEVC for still image (HEIF)