Skip to content

MobinaPoulaei/DL_HW1_SUT

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 

Repository files navigation

Deep Learning Course - Sharif University of Technology

Instructor: Dr. Mahdieh Soleymani

This repository contains my solutions and code for the first homework assignment of the DL course, covers fundamental concepts of deep learning, from theoretical foundations to practical implementation using PyTorch.


Theoretical Section

The theoretical portion of this homework covers fundamental mathematical concepts crucial to deep learning. My handwritten answers are submitted in a single PDF file as required by the course. The topics covered are:

  • Matrix Differentiation: Derivations for various matrix expressions.
  • Backpropagation: Step-by-step calculation of gradients for a custom neural network architecture.
  • Optimization: Analysis of the Backtracking Line Search and Adam optimization algorithms.
  • Regularization: Concepts related to Lipschitz continuity and the role of Weight Decay.

Practical Section

This part of the homework involves hands-on implementation using PyTorch. The code is organized into separate files for each question:

  • Basics: An introduction to PyTorch, where I implemented a simple classification task. The required files are Basics.ipynb and pytorch_basic.py.
  • NN Scratch: Building a complete neural network using custom modules without relying on PyTorch's built-in nn library. The utils folder, which was provided, is essential for running this part of the code.
  • Optimization: A practical assignment on optimization algorithms that was submitted in person.
  • Lazy Gradient: An assignment exploring memory challenges when working with neural networks.

About

This repository contains my solutions and code for the first homework assignment of the Deep Learning course at Sharif University of Technology.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors