报告题目：Quantum Reinforcement Learning
报告人：Houghton College Prof. Wei Hu
报告主要内容：Machine learning can be generally categorized into three classes, supervised, unsupervised, and reinforcement learning. It has played an important role in the analysis of big data and artificial intelligence today. The year of 2016 has witnessed the single greatest AI achievement in human history: AlphaGo, a computer program developed by Google which defeated the best human Go player, a feat no one could ever imagine even just a few years ago. The success of AlphoGo is based on the recent development in machine learning such as deep neural network, convolution network, and reinforcement learning.
A quick question after the excitement of AlphaGo phenomenon is what’s next? There is no doubt that computing power is still one of the main bottlenecks of AI advancement, as such turning our attention to quantum computing and quantum algorithms is our natural choice. Based on quantum mechanics, quantum computing is a very different computing paradigm from its classical counterpart as quantum states could be manipulated in parallel.
In this talk, I plan to introduce a famous quantum search algorithm, Grover search algorithm, and explain how it could be applied to reinforcement learning. Experiments have shown that even on the classical computers, such application can dramatically reduce the time to train a machine learning model and the resultant model is more robust to nose in data, a feature very desirable in the world of machine learning. Therefore we do not need to wait for a general purpose quantum computer to become available on our desk, we can take advantage of its power today by using our classical computers.