ECE57000
Artificial Intelligence
Fall 2018

General Information

Lectures: M,W,F 12:30pm - 1:20pm, PHYS 203.

Professor: Jeffrey Mark Siskind, EE313E, 765/496-3197, qobi@purdue.edu
Office Hours: M,W 5:00pm - 6:00pm

  • nearest neighbor classifiers
  • k-means clustering
  • the EM algorithm
  • hidden Markov models
  • the Viterbi algoritm
  • the forward/backward algorithm
  • the Baum-Welch reestimation procedure
  • probabilistic context-free grammars
  • forward and reverse mode automatic differentiation
  • neural networks
  • multilayer perceptrons
  • backpropagation
  • convolutional neural networks
  • object classification
  • object detection
  • recurrent neural networks
  • LSTMs
  • image captioning
  • deep reinforcement learning
  • constraint satisfaction problems
  • graphical models
  • probabilistic programming
    The course requirements and grading for ECE57000 during the fall 2018 semester will be different than in previous years. There will be no midterm and final exams. There will be between three and six problem sets. The exact number and due dates will be determined over the course of the semester. These will count for 30% of the grade in the course. The remaining 70% of the grade in the course will be based on the following requirements which will take the place of the exams. The grade in this course will be a letter grade with no + or -.

    Students will be required to select and read three recent conference or journal papers in the fields of AI, computer vision, robotics, natural language processing, cognitive science, neuroscience, or machine learning. Nominally, the papers should have been published within the last three years in one of the following venues: AAAI, IJCAI, AIJ, JAIR, CVPR, ICCV, ECCV, PAMI, IJCV, ICRA, IROS, RSS, TRO, ACL, EACL, NAACL, CL, TACL, Cognition, CogSci, Nature Neuroscience, Neuron, J Neuroscience, Neuroimage, NIPS, ICML, UAI, AISTATS, or JMLR.

    You are welcome to read more than three. Your selection of papers must be approved by me. I may be willing to accept papers from venues other than the above and/or older than three years old. But all paper selections, whether or not they meet the above criteria must be approved by me. You must submit your paper selections by 5:00pm Friday 7 September 2018 with the following information:

    (a) a BibTeX entry for each of the papers that you would like to read. For a journal paper this should contain (at least) the paper title, authors, journal, volume, year, and pages. For a conference paper this should contain (at least) the paper title, authors, conference, year, and pages. (b) a URL containing a pdf file of each paper

    Paper selections will be submitted through Blackboard using a mechanism that will be announced later in the semester.

    Please note that you should not send me the paper itself. Not even the abstract.

    If you intend to select papers that are older than three years old or from a venue other than listed above, I suggest that you discuss this with me prior to Friday 7 September 2018.

    You will be required to read all three papers and implement and evaluate the ideas from (at least) one of the papers. Thus (at least) one of the papers that you select should contain material that is suitable for implementation. The implementation must be nontrivial. A good guideline is that the implementation should be at least four pages of code. This is not a strict guideline. Ultimately, I will determine whether or not the implementation meets the non-triviality requirement. The implementation can be in any programming language that you choose, though obviously lower level languages may require much more code to implement the same functionality that could be achieved with less code in a higher level language. You must also conduct a substantive evaluation of your implementation to determine how well it solves the intended problem. Ideally, you should replicate the experiments presented in the paper but I will not require this.
    You will be required to write a six page paper in LaTeX meeting the typesetting conventions of AAAI. Approximately three pages of this paper should be a substantive critique of the three papers that you have read. And approximately three pages of this paper should be a description of your implementation and evaluation of the material from one of the papers. The term paper must be submitted by 5:00pm Friday 16 November 2018.

    Term papers will be submitted through Blackboard using a mechanism that will be announced later in the semester.

    The term papers will be reviewed by other students in the class. Like all conferences, this process will be double blind: reviewers will not know the identity of authors and vice versa. To support this, like all conferences, you should NOT put your name on the term paper submission. In place of your name, you should put your PUID. Also like conferences, reviews will be confidential. The only person who will be privy to the reviews will be the reviewer, the instructor, and the author.

    I will assign each term paper to five students to read. Each student will be required to read five student term papers (other than their own) and prepare conference-style reviews, primarily indicating clarity and the quality of the implementation effort. Peer reviews will be due by 5:00pm Friday 7 December 2018. The exact format for the review will be determined later in the semester.

    Reviews will be submitted through Blackboard using a mechanism that will be announced later in the semester.

    You will also be required to make a 10 minute conference-like presentation in class during the last six weeks of classes. The exact schedule will be determined later in the semester. But to accommodate all students in class, presentations will start on Monday 15 October 2018. So you should be prepared to give your presentation at any time after that. This is necessarily before the due date for paper submissions and the due date for reviews. This presentation should cover both a summary of the three papers that you have read, a summary of your critique of those papers, and a description of your implementation and evaluation. The presentation must be given from a laptop or the computer that is builtin to the lectern in the lecture hall. You can use whatever tools you wish to prepare your presentation (i.e. LaTeX/beamer or PowerPoint). On the day that you are scheduled to give you presentation, you should arrive in class early to make sure that you presentation setup works and that you are prepared to give your presentation in the allotted time slot.

    Students will be required to attend all student conference-like presentations. Attendance will be take in the student conference-like presentations. A passing grade will require that you not miss more than two days of class that contain the student conference-like presentations.
    EE570 | ECE | ECN | Purdue University