Intro here

• Instructor: Dr Anthony Maida. Oliver Room 355.

• email: maida@louisiana.edu

• Class meeting time: Tues/Thurs 4:00pm - 5:15pm.

• Dr Maida Office Hours: Tues/Thurs 3:00 pm - 4:00 pm.

•                                         Mon/Wed 2:00 pm - 4:00 pm.

• Class Room: 116 Oliver Hall.

• GTA: Sai Venkatesh Chilukoti, sai-venkatesh.chilukoti@louisiana.edu

• GTA Office hours: starting Monday, Feb 5.

• 12 noon - 2PM, Monday through Thursday, Room 356.

• List of assignments and due dates:

  • First programming assignment: Linear regression. Due Feb 20, 11:59PM. Instructions are at Week 4, Entry 7. Turnitin portal is just above instructions at Item 6.

  • First quiz (take home): Derivations for gradient and cross-entropy. Due in class, Feb 27. Instructions are at Week 5, Item 7.

  • Second programming assignment: House Price Prediction. Due March 5, 11:59PM. Instructions are at the end of Week 6, Item 6. Turnitin portal is just above instructions at Item 5.

  • Second quiz is posted on Moodle (Week 9, Entry 6). It will be due 11:59PM, March 14, Thursday. Sample questions are found on Week 9, Item 5.

  • Midterm exam, in class, Tuesday, March 19. The exam syllabus is posted in Week 7, Item 1.

  • Sample questions for the midterm are posted under Week 10, Item 5.

  • Short assignment posted. See Week 11, item 3. Asks you to get started with LaTeX.

To see what's out there in the AI world, here are some useful or interesting web sites:

• pytorch.org Offers excellent tutorials on PyTorch.

• kaggle.com  Offers tutorials, datasets, and competitions.

• huggingface.co Offers pretrained models, source code and datasets.

• chat.openai.com  An extremely high-powered chatbot.

• openai.com/dall-e-2/ An image generator.

• stability.ai/blog/stable-diffusion-public-release  Another image generator.

• colab.research.google.com   Online Python notebook environment w/ GPU support.

1: Course information, textbook, and syllabus

File

Please click to view and download.

Updated: March 3, 2024.

2: Creating Virtual Env for PyTorch

Updated: 1/17/24

This is one way to install PyTorch.

It is recommend that you install PyTorch in a virtual environment. This creates a sandbox for your installation which does not interact with the other software on your computer.

Most people install Python and PyTorch using Anaconda or Miniconda. Here is the link for Miniconda: https://docs.conda.io/projects/miniconda/en/latest/

Alternatively, you can use Python package installation tools. Here is link to documentation on that: https://packaging.python.org/en/latest/tutorials/installing-packages/

The above download illustrates a way to do the latter method for Mac OS. Documentation for other platforms is readily available.

3: Installing d2l APIs

Updated: 

File

The Dive into Deep Learning text adds its own set of APIs to PyTorch for teaching purposes. Download this handout for instructions to install the D2L APIs on your computer.

Also, the book explains how to do a lot of the programming using notebooks in Google colab available at:

https://colab.research.google.com

For colab, install d2l with the command below.

!pip install d2l==1.0.3

Later chapters in the book depend on software from earlier chapters, so essentially you have memorize the code base to get through the book.

Test if d2l APIs are installed.

File

Added: 1/17/2024

Run this code, taken from chapter 3, to test of the d2l APIs are installed on your computer. It should print a tensor and plot three Gaussian distributions.

This code was used in the spyder editor and development environment. If it works you, you should also be able to use jupyter notebooks on your machine, by typing "jupyter notebook" in the terminal.

3a: Look at source code for d2l framework. File: torch.py

Optional.

Click to download d2l.py file. It is worth looking at to see how utilities have been defined. Do not modify this file.

In the textbook, whenever you see a Python definition annotated with the #@save symbol, it has already been defined in the d2l.py file.

Note: To install the software on your computer, click the notebooks tab on the top of the d2l.ai web page. You will download a folder called d2l-en. It has a subdirectory pytorch. WIthin that directory, copy the subdirectory d2l to your .ipython hidden folder. This installs the d2l utilities on your a machine. Other folders in the directory contain ipython notebooks for each chapter in the book. You can upload a notebook to colab.research.google.com and run it.

3b: Accessing CACS GPU servers

File

These are instructions by Dr Minvielle and accessing GPU servers in CACS. The relevant instructions start on slide 14.

4: Basic tutorial on numpy

File

"numpy" stands for numeric Python. It's primary data type is an "ndarray" which stands for n-dimensional array. The deep learning frameworks use tensors, which are modeled after ndarrays.  Learning numpy prepares you for using the DL frameworks.

5: Short Brain Overview

File

Neural networks in DL were originally inspired by the connectivity patterns of biological neurons in the brain. This presentation gives you a sense of how part of the brain --- the neocortex --- is organized.

For study purposes, just look at the first 13 slides.

6: Intro to Feedforward Networks

File

The main purpose of this presentation is to give you a feeling for how multi-layer neural networks can be represented compactly of stacks of matrices.

Study slides 1-6, 9, and 15-16.

Assignment: 

Read Chapter 1. And look at  the websites below.