عبدالكريم الخطيب

Harvard CS197: AI Research Experiences & Deep Learning Guide

3 دقائق للقراءة

في هذه الصفحة
  1. Table of contents
  2. Reviews
  3. Further Reading

Table of contents

  • This course (Harvard CS197, also written CS 197) consists of 21 quick lectures that include valuable experiences and important tips for anyone interested in the field of scientific research, especially deep learning. The course can be completed in approximately one or two days. The title is not very precise, and the content of the lectures ranges from about 8 to 26 pages each. Some topics may not be directly related to the specifics of scientific research, but they are generally very helpful lectures for beginners in the field.

Reviews

Lecture 1: Exciting Advances with AI Language Models Content : Interact with language models like GPT-3’s text completion and use Codex’s code generation abilities feedback : ⭐ (1/5)

Lecture 2: The Zen of python Content : vscode,git,conad,linting and Debugging. feedback: feedback : ⭐ (1/5)

Lecture 3: Reading AI Research papers Content :

  1. Conduct a literature search to identify papers relevant to a topic of interest
  2. Difference between Reading Wide and Reading deep and how to balance between them
  3. How to use Google Scholar and paper with code feedback : ⭐⭐⭐⭐⭐ (5/5)

Lecture 4: In-Tune with Jazz Hands Content:

  1. quick intro into huggingface
  2. Tokenization
  3. Causal language modeling (CLM) feedback : ⭐⭐⭐⭐ (4/5)

Lecture 5: Lightning McTorch Content :

  1. Fine-tuning A vision Transformer
  2. Intro to pytorch lightning (Lightning)
  3. Data Loading
  4. How to Build a Neural net Module with lightning and how lightning modules work feedback : ⭐⭐⭐⭐ (4/5)

Lecture 6 & 7: Moonwalking with Pytorch Content :

  1. Pytorch Exercises
  2. Tensors
  3. Autograd and neural networks feedback : ⭐ (1/5)

Lecture 8 & 9: Experiment Organization Spakrs Joy Content :

  1. Weight and Biases
  2. Hyperparameter Search
  3. Hydra feedback : ⭐⭐⭐⭐ (4/5)

Lecture 10 & 11 : I Dreamed a Dream Content

  1. Identifying Gaps in A Research Paper
    1. CLIP and CheXzero
  2. Generating Ideas for Building a Research Paper
  3. Iterating on your research ideas feedback : ⭐⭐⭐⭐⭐ (5/5)

Lecture 12 & 13 : Today Was a Fairytale

  1. how to deconstruct the elements of a research paper and their sequence
  2. Resulting template that you can use as a general example feedback : ⭐⭐⭐⭐ (4/5)

Lecture 14 & 15: Deep Learning on Cloud Nine didn’t complete it 🙃🙃🙃

Lecture 16 & 17:Make your dreams come tuned Content

  1. high level use of Stable Diffusion using a Dreambooth template
  2. Use AWS to accelerate the training of Stable Diffusion models with GPUs
  3. HF Accelerator feedback : ⭐⭐ (2/5)

Lecture 18 : Research Productivity Power-Ups Content

  1. How update meetings and working sessions
  2. organizing your efforts on a project
  3. what is technical dept and examples on it feedback : ⭐⭐⭐⭐ (4/5)

Lecture 19 :The AI Ninja Content

  1. How to make Steady Progress
  2. Some Research Skills
  3. Discussion Questions feedback : ⭐⭐ (2/5) I found that Colah’ blog content about research is better in the context and offers a great details

Lecture 20: Bejeweled ⭐⭐⭐⭐⭐(5/5)

  1. how to make a slides to improve your research talk
  2. Assertion Evidence Approach feedback : ⭐⭐ (2/5) This is great related talk from MIT about this topic How to speak ⭐⭐⭐⭐⭐(5/5)

Lecture 21 : Model Showdown Content

  1. Statistical Testing feedback : ⭐⭐ ⭐(3/5)

Further Reading

If you found this review of AI research methodologies useful, you might also be interested in:

خريطة الروابط