Lossless high entropy compression algorithm


  • Machine Learning

  • Signal Processing

  • Neural Networks


Mentors :

  • Anubhav Agarwal

Mentees :

  • Aman Singh
  • Raj Aryan Agarwal
  • Gaurav P
  • Tirthankar Adhikari
  • Ojas Gramopadhye


Implement your own sequence neural net

Ever dreamed of making the compression algorithm from Pied Piper featured in the popular Silicon Valley series? Here is your chance! The project aims at implementing a potentially lossless algorithm which, unlike others, can work even on high entropy data. The project will involve basic knowledge of Discrete Time Fourier Transform, signal processing. However, a major portion would involve developing expertise in neural networks with focus on RNNs, LSTMs, etc. You would implement your own sequence to sequence neural net aimed at performing a particular task which is essential to the working of compression algorithm.

Tentative Timeline :

Week Work
Week 1,2,3 Developing fundamentals in python,machine learning and neural networks through popular sources like coursera, udacity.
Week 4,5 Learning basics of RNNs and LSTMs and their in-depth implementation in tensorflow 2.0
Week 6,7 Trying out various models and getting your hands dirty while aiming to achieve the objective
Week 8,9 If above task is complete, packaging everything into an end-to-end pipeline