Logistics

Instructors:

TAs:

  • Yashovardhan Chaturvedi (yashovac at andrew.cmu.edu). Office: GHC 6708 Fri 1:30 - 2:30 PM
  • Liang Zhang (liangz1 at andrew.cmu.du). Office: GHC 5417 Tue 1:00 - 2:00 PM

Meetings: Tue and Thu 10:30 - 11:50 at GHC 4215

    Piazza: https://piazza.com/cmu/spring2019/cs11695

    This is a hands-on implementation course focusing on building deep learning networks using Python and Tensorflow. You will develop the following skills:

    • Program supervised learning, semi- and un-supervised learning, deep reinforcement learning, probabilistic networks
    • Apply deep learning models, including multilayer perceptrons, convolutional neural networks, recurrent neural networks, encoder-decoder and probabilistic models
    • Construct computer vision, machine translation, game playing systems and others.

    TOPICS

    • Basic Numpy, Tensorflow
    • Feed-forward Network
    • Convolutional Neural Network and Image Classification
    • Recurrent Neural Network, Sequence-to-Sequence and Machine Translation
    • Reinforcement Learning
    • Some Advanced Topics
      • Variational AutoEncoders
      • Generative Adversarial Networks
      • Object Detection
      • Text Style Transfer
      • Optional: MCMC, Gaussian Processes, Neural Architecture Search

    PREREQUISITES

    • Probability and Statistics: PDF, CDF, Basic Distributions, Basic Bayesian Statistics
    • Linear Algebra: Vector, Matrix, Eigen Value Decomposition, Eigen Vectors
    • Multivariate Calculus: Integrals, Gradients, Partial Derivaties, Hessians
    • Proficient Python

    Assessment

      1. There is NO mid-term or final exam. You will be graded based on weekly quizzes and 3 big assignments, as follows:
    • (25%) Weekly quizzes: there will be a 15-min test by the end of each Thursday’s lecture which will cover the knowledge of the previous week. There will be multiple-choice questions and in some cases, some short explanation is also expected. Two worst quizzes will be dropped.
    • (75%) Coding assignments: there are 3 assignments which require you to work individually. The code will be submitted to the shared TA email and will be graded individually.
      2. NOTE: You only have 1 week to dispute or make a regrade request for any quizz/homework. After that, the grade is unchangable.
      3. GRACE DAY: Each coding assignment has only 1 grace day period. After that
      • 1 late days: - 30%
      • 2 late days: - 60%
      • And no more, i.e. ZERO otherwise.

    RESOURCES

      For textbooks, this course requires no official textbooks but we recommend you to read the following texts:
      Furthermore, the instructors and TAs will update useful links on Piazza to assist you more on learning the theories and implementation.
      For compute resources, we will provide all students enough Amazon AWS credits. The TAs will instruct you on how to use AWS and Google Colab resources which also provide you abundant free compute power for testing the codes.

    PLAGIARISM

      All work in this course, including both quizzes and assignments, must be your own individual work. That is, you must work alone on quizzes and assignments. If you include material from another source in any work, you MUST provide a citation to the source.
      Violations of the Academic Integrity Policy are taken very seriously and the instructors MUST report any violations as provided by policy.
    Adopted from Start Bootstrap.