Kewei Qu

Mentor: Professor Douglas Blank

Department: Computer Science

Exploration of Deep Learning in Robotics

Deep learning, a new phrase coined in 2006, is a sub area of Machine Learning methods that aims at true Artificial Intelligence (AI). Instead of focusing on producing “human” results by machine based on large amount of sample data, deep learning hopes to understand how human brain functions and produce such brain “learning” without human supervision in machines. This summer, I will explore core methodology of deep learning networks: unsupervised training session and its applications in speech recognition.


Description: Macintosh HD:private:var:folders:nf:fwv2qkd911vbmr_f53n8l9k00000gn:T:TemporaryItems:imgres.jpg

The focus of the project at this stage is to explore algorithms of deep learning and how it applies to robotics. Our goal is to program a robot with relatively huge autonomy that can decide on its own what to do as it experiences “life,” e.g. how to evaluate “boringness,” what motivate it to learn new things? We will start off with simple neural network architecture: single perceptron and gradually move towards more complex architectures that are non-linear.