CSCE 790 Machine Learning
Spring 2018

Course Description: In this course we will study the latest methods in Machine Learning as they apply to the field of robotics. In particular we will study: Reinforcement learning, Gaussian Processes, Deep Learning, and Deep Reinforcement Learning. The course will be divided into studying background material and state of the art papers. Course learning outcomes:
  • Develop necessary research skills
  • Conduct a literature review on a selected topic
  • Present a scientific paper
  • Summarize the content of a presentation
  • Develop the ability to work with robotic systems
  • Develop the ability to work with learning frameworks
Prerequisites: There are no prerequisite courses. However, students are expected to have strong software development skills. Projects are likely to involve some combination of C++, Python, bash, Linux, ROS, LATEX, and other tools as needed. Experience in machine learning and robotics is recommended, but strong students may be able to acquire the necessary background on-the-fly.

Instructor

Evaluation:

  • Assigniments (5) 10% each: 50%
  • Presentation Summary 10%
  • Class participation 10%
  • Presentations 30%
 

Student Presentations


Syllabus


Resources

Paper
Resources Presenter Presentation Date Remarks
POMDP presentation Sharaf Malebary
Coastal Navigation
DroNet Logan Murray 2/15/2018
Gaussian Processes
Gaussian process modeling of large-scale terrain Presentation JEREMY A DAY
"Safe Exploration in MDP" by Teodor Mihai Moldovan and Pieter Abbeel Nare Karapetyan
Deep Learning Part I MD Modasshir
Human-level control through deep reinforcement learning Presentation Bharat Joshi 4/5/2018
GONet: A Semi-Supervised Deep Learning Approach For Traversability Estimation Project website Nawras Alkassab
Persistent self-supervised learning principle: from stereo to monocular vision for obstacle avoidance Presentation Michail Kalaitzakis 4/10/2018
Learning by Demons tration for Motion Planning of Upper-Limb Exoskeletons Presentation Ivan Panchenko 4/12/2018
Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks Presentation Ying Meng 4/17/2018
DPC-Net: Deep Pose Correction for Visual Localization Hussein Almulla 4/19/2018
DeepVO: Towards End-to-End Visual Odometry with Deep Recurrent Convolutional Neural Networks Ahmed H Saaudi 4/24/2018
Other Papers

Lecture notes are posted here

Gaussian Processes for Machine Learning, Carl Edward Rasmussen and Chris Williams, the MIT Press, 2006, online version

Machine Learning: a Probabilistic Perspective. Kevin P. Murphy, MIT Press.

Reinforcement Learning, An Introduction. Richard S. Sutton and Andrew G. Barto. MIT Press

A Gentle Introduction to ROS

Assignments

Description
code
worlds/data files
Due date
Assignment 1: Intro to ROS and Random Walk
08 Feb. 2018
Assignment 2: Bibliography Search

22 Feb. 2018
Assignment 3: Programming Assignment using a CNN notebook code

Assignment 4: Programming Assignment using GP Courtesy of Professor D. Meger from McGill University

Assignment 5: Submit a report summarizing the class