Hysen Labs
Open-source project
floodsung/Deep-Learning-Papers-Reading-Roadmap avatar
floodsung

Deep-Learning-Papers-Reading-Roadmap

Deep Learning papers reading roadmap for anyone who are eager to learn this amazing tech!

39,547 stars7,269 forksPythonLicense varies
01
DEEP OPEN-SOURCE ANALYSIS

Where to start reading deep learning papers, ranked by importance

The roadmap orders the field from outline to detail and from old papers to state of the art. It opens with a book and a Nature survey, then walks through milestone and speech papers, each entry tagged with a star rating.

02
DEEP OPEN-SOURCE ANALYSIS

Who the roadmap is for

The README addresses anyone eager to learn deep learning and names the first problem: picking a starting paper. It structures the path three ways at once, from outline to detail, from old to state of the art, and from generic to specific areas, with entries grouped by topic. Each listed paper carries a title, its authors, and a star rating for importance. Python is recorded as the repository's primary language.

03
DEEP OPEN-SOURCE ANALYSIS

A book, then a survey

The first recommended stop is the Deep Learning book by Bengio, Goodfellow, and Courville, which the roadmap calls the deep learning bible. A survey by LeCun, Bengio, and Hinton published in Nature follows as the second key starting point, referred to in the file as the three giants survey. Together the two give a reader a fixed point of entry before anything newer is attempted.

04
DEEP OPEN-SOURCE ANALYSIS

Milestone papers

The milestone list starts with the deep belief network work by Hinton, Osindero, and Teh, then the dimensionality reduction paper by Hinton and Salakhutdinov. The ImageNet evolution is traced through AlexNet, VGGNet, GoogLeNet, and ResNet, which the README marks as deep learning breakthroughs. Each entry keeps the same shape: paper title, authors, and an importance rating, so the ordering can be skimmed fast.

05
DEEP OPEN-SOURCE ANALYSIS

The speech recognition line

The speech section follows the field as it moved: deep neural networks for acoustic modeling, speech recognition with deep recurrent neural networks, end to end speech recognition, and Deep Speech 2. These four entries show the same format as the rest of the roadmap, with titles, authors, and star ratings. The progression reads like a compressed history of one research thread.

06
DEEP OPEN-SOURCE ANALYSIS

What the rating system carries

The star ratings are the organizing device of the whole list. A newcomer can treat them as a priority queue and work down the ranked entries, while the topic grouping keeps related papers together. The roadmap makes no claim to be complete; it is a curated path with a clear starting point and a clear order, and it stays a reading guide rather than a summary of the papers themselves.

07
DEEP OPEN-SOURCE ANALYSIS

Editorial conclusion

None of this substitutes for reading the papers, and the roadmap does not pretend otherwise. It exists to answer the one question every newcomer asks: which paper first.

08
DEEP OPEN-SOURCE ANALYSIS

Official sources

09
Community notes

Community notes