Skip to content

This project develops a deep learning model for image classification, focusing on accuracy, uncertainty, and robustness. A custom CNN was built using the Adam optimizer and cross-entropy loss. Uncertainty metrics (least confidence, entropy) and diversity metrics (cosine similarity, Kullback-Leibler divergence) were used to evaluate performance.

Notifications You must be signed in to change notification settings

n1kh1le5hp/WINTER-PROJECT-ACTIVE-LEARNING

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

39 Commits
 
 
 
 

Repository files navigation

Resources

Week 0: Installation Guides

Python 3

Anaconda


Week 1: Getting Started Tutorials (Mandatory)

Git and GitHub

Markdown

Jupyter Notebooks

Python and its Libraries

Detailed Courses (Recommended)

Textual Resources

  • W3Schools - Sections on Python, NumPy, Pandas, Scikit-learn, and Matplotlib
  • GeeksforGeeks - Sections on Python, NumPy, Pandas, Scikit-learn, and Matplotlib

Video Resources

Session Link:

https://drive.google.com/file/d/1GWDBJ56N2etZZhQhrXXRs5LxGdjYxbXQ/view?usp=drivesdk

Assignment 1:

  • Complete 10-15 questions manipulating datasets and working with the libraries listed above.

Week 2: Machine Learning Problems - Regression and Classification

Regression

Classification

PyTorch Basics

Assignment 2:

  • Implement a PyTorch classifier on a basic dataset.

Week 3: Active Learning Methodologies

Research Papers on Query Strategies

Assignment 3:

  • Implement Active Learning using the query strategies discussed in the papers and evaluate the performance.

About

This project develops a deep learning model for image classification, focusing on accuracy, uncertainty, and robustness. A custom CNN was built using the Adam optimizer and cross-entropy loss. Uncertainty metrics (least confidence, entropy) and diversity metrics (cosine similarity, Kullback-Leibler divergence) were used to evaluate performance.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 100.0%