AI-Lab
- Machine Learning
- Deep Learning (e.g. vision, sequence and reinforcement learning)
- Advanced Techniques (e.g. GAN, Zero-Shot Learning, Transformer, Multi-Agent Learning etc.)
Now, I'm conducting research on Reinforcement Learning
The best way to get deeper into AI technology is to get hands-on with it. In 2018-2020, I have an exciting plan to conduct a dozen experiments with numerous and diverse deep learning applications based on Computer Vision (CV), Natural Language Processing (NLP) and Reinforcement Learning (RL) technologies. Through learning-by-doing, the following is the list of applications I envision:
CV Applications
1. Object Classification
2. Object Detection
3. Real-time Object Detection
4. Semantic Segmentation
5. Instance Segmentation
6. Human Pose Detection
7. Visual Question Answering
NLP Applications
1. Machine Translation System
2. Sentiment Analysis
3. Text Summarization
4. Topic Modeling
5. Chatbot
6. Image Captioning
7. Speech Recognition
RL Applications
1. Dynamic Programming Method for MDPs
2. Monte Carlo Method
3. Temporal-Difference Method (Sarsa, Sarsamax, Expected Sarsa)
4. Value-Based Method (DQN, Double-DQN, PER-DQN, Dueling-DQN, Noisy-DQN, Distributional-DQN, Rainbow-DQN)
5. Policy-Based Method (Reinforce, TRPO, PPO)
6. Actor-Critic Method (A2C/A3C, GAE, DDPG)
7. Multi-Agent Method (MADDPG, MFMARL)
Robotics Applications
The time will come soon.
Dependencies
This lab requires Python 3.7.3 and the following Python libraries installed:
- Basic Libraries: NumPy, Matplotlib
- Domain-specific Libraries: OpenCV, NLTK, Gym
- Deep-learning Frameworks: Keras, PyTorch, TensorFlow, ReNom