everything.study
15.773

Hands-on Deep Learning

3-0-3 units · Graduate

Fast-paced introduction to deep learning, the engine behind modern artificial intelligence and generative AI, with an emphasis on developing a practical understanding of how to build models to solve complex problems involving unstructured data. Topics include: the basics of deep neural networks and how to set up and train them, transformers for natural language processing and computer vision, large language models (such as ChatGPT), and text-to-image and multi-modal models. Prior familiarity with Python and fundamental machine learning concepts (such as training/validation/testing, overfitting/underfitting, and regularization) required.

Spring 2024

Topics

  • Introduction to neural networks and deep learning (motivation, smart representations, layers, activations, architectures); training deep NNs (loss functions, gradient descent, regularization)
  • Training deep NNs (continued); introduction to Keras/Tensorflow (tensors, functional API, training, evaluation); application to tabular data
  • Deep learning for computer vision—building convolutional neural networks from scratch
  • Deep learning for computer vision—transfer learning and fine-tuning; introduction to HuggingFace
  • Deep learning for natural language—the basics (tokenization, n-grams, bag-of-words)
  • Deep learning for natural language—embeddings
  • Deep learning for natural language—transformers
  • Deep learning for natural language—transformers, self-supervised learning
  • Generative AI—large language models (LLMs) and retrieval augmented generation (RAG)
  • Generative AI—adapting LLMs with parameter-efficient fine-tuning
  • Generative AI—text-to-image models
  • Project presentations

Readings

Prerequisites & course information

Prereq: 15.060 , 15.071 , 15.072 , 15.774 , or permission of instructor G (Spring; first half of term) 3-0-3 units Credit cannot also be received for 15.776

Sources