Overview

You can see this repo as a consolidation of reading groups which I do between me and myself :)

Why this repo

Keeping the pace of current Deep Learning Research is almost impossible and even focusing on a specific field, like Computer Vision, is extremely hard because of the new papers rate (which by the way I think it is a great thing, it makes us living in exciting times), their hard to understand content and the fact code is not so often accompanying the paper

This is specifically true if you are student as, in my opinion, the gap between the educational material (books, courses, ...) and the research material today is quite big.

The goal of this repo is to try helping students to make the quantum leap to get from the students side to the researchers side with the following methodology

• Selecting relevant or interesting papers
• Providing papers summaries
• Adding code snippets implementing specific parts
• Diving a bit more into theory and math when necessary (research papers tend to be vague on this kind of things)

How to CONTRIBUTE

As explained in Contribute.md contribution are absolutely welcome

• Open issues to

  • bugfix: notify me about bugs, typos, ...
  • features: suggest relevant papers, discuss content, ...

• Submit PR with new summaries, implementations, ...

Feel free to DM me on Twitter at @NicolaBernini to discuss anything

How to View the content

The content is typically Markdown because it is very light but as you know GitHub does not support server side math rendering so I suggest to use some Browser Plugin like

• Tax All The Things

• MathJax Plugin for GitHub

Progress

What's new

Recent

Date	Topic	Content
2020-10-11	RL Framework - Visual see above	Added in the ReinforcementLearning section
2020-08-24	Focusing on the Deep Learning Code Development Challenges	Added Challenges in Deep Learning Code Development
2020-08-17	From Papers to Code: implementing Yolo v1 from the paper	Showing how to read a paper defining an architecture and how to implement the architecture directly in Pytorch
2020-08-16	Added Kanban Secion	In this Kanban I am going to collect interesting bookmarks related to papers and blog posts to read
2020-08-15	Added WorkingOn Section where I am happy to share papers and blog posts which I am currently reading	Started reading this interesting post from Adam Kosiorek Blog about ML of Sets
2020-08-14	Added section about Domain Adaptation with good first paper analysis	Universal Domain Adaptation through Self Supervision
2020-07-31	Added Analysis of a New Paper	Analysis of An Optimistic Perspective on Offline Reinforcement Learning - Gist
2020-07-10	Added anew Section on GNN	Started the new Section about GNN with a paper summary Discovering Symbolic Models from Deep Learning with Inductive Biases Also available on Colab as Discovering Symbolic Models from Deep Learning with Inductive Biases
2020-06-21	Disentangled Representation	Commented Disentangling by Factorising on Scirate
2020-04-18	Causality and Machine Learning	Added related comments here https://github.com/NicolaBernini/PapersAnalysis/issues/33
2020-02-24	Domain Adaptation	Paper Read - Universal Domain Adaptation through Self Supervision https://github.com/NicolaBernini/PapersAnalysis/issues/32

Full List of Work in Progress

https://github.com/NicolaBernini/PapersAnalysis/issues

By Topic

Point Cloud Processing

Representation Learning

Reinforcement Learning

To be fixed list

2020-06-21

Commented Disentangling by Factorising on Scirate

2019-11-12

• Updates in Reinforcement Learning

2019-10-19

Added ~~Reinforcement Learning, Fast and Slow notes~~

• Removed the PDF with a Collaborative Analysis Doc on Reinforcement Learning, fast and slow with comments open

2019-06-28

Added Keras in depth tutorial - MNIST and CNN focused less on the how-to and more on the why and what happens under the hood

• feel free to comment on the Medium Version

2019-06-16

Added Summary and Notes - Work in Progress about Andrew Zisserman Talk about Self Supervised Learning - 2018

2019-06-15

Added Summary and Notes - Work in Progress about Daniel Roy's talk at ICML 2019

2019-06-12

Added a Paper Readthrough Issue related to Reinforcement Learning, Fast and Slow paper from DeepMind

2019-05-17

Added a PR about Reservoir Computing started with some Basic Elements elements

2019-05-13

Added PR related to a new Paper Summary regarding CNN Debugging with Heatmaps

Added a Paper Readthrough Issue related to Quantum reservoir processing : very interesting Quantum Machine Learning paradigm

2019-05-07

Added a PR about Geometric Deep Learning starting with Gauge Equivariant CNN paper summary

Added a PR about Fake News starting with a Paper Summary

2019-05-05

Added a PR about Statistics and the first summary regards Variational Bayesian Monte Carlo also published as Kaggle Kernel

2019-04-29

Added a PR to add D2 Net Summary

2019-04-28

Added a PR to define the CNN in Computer Vision - Section starting with a Paper Summary

Published on Medium

• Paper Summary (Serial): Benchmarking Classic and Learned Navigation in Complex 3D Environments

Sparse Topics

Deep Learning Theory

• Deep Learning without Poor Local Minima

Reinforcement Learning

• Reinforcement Learning, Fast and Slow

• Q Learning

News

• Yoshua Bengio - Gary Marcus - AI Debate

Work in Progress

Body of Knowledge

• Deep Learning - Body of Knowledge

• Generative Models - Body of Knowledge

Papers Readthrough

• Complex YOLO