Reinforcement Learning (RL)

Response to 2016/17 survey feedback for RL.

We have taken note of the feedback from students in this course survey, and have the following comments in response.

  • Firstly, we are pleased to hear that many students agree with the instructor(s) on the timeliness of the topics being covered, and the importance of RL within the AI curriculum.
  • Many students have noted that this was a difficult course due to many organisational issues, including the lack of tutorials outside the main series of lectures (which in the 2016/17 was owing to to insufficient resources being available with respect to the number of registered students). We can say that arrangements have already been made with the teaching office regarding tutorials and lecture recordings in subsequent offerings of the course.
  • Many students liked the assignments based on the Atari Learning Environment, but felt that it was a lot of work considering the marks allocated to it. We have already increased the fraction of final mark attributed to the coursework in subsequent offerings of this course. We will also carefully consider how to better scope the assignments so that students continue to benefit from this (or similar) environment, but do not find the time requirements to be unreasonable.
  • We acknowledge the feedback that the range of topics covered makes the course more fast paced than necessary. We will attempt to calibrate the pace better in future offerings, by restricting the special topics and allocating more time to the basic concepts.

Subramanian Ramamoorthy, September 2017