What it does This is an “awesome list” — essentially a human-curated index of research papers and code repositories focused on neural architecture search (NAS) and hyper-parameter optimization. It catalogs work from 2015–2019, organized by technique: reinforcement learning, evolutionary algorithms, and miscellaneous approaches, plus a separate section on hyper-parameter search methods.

The interesting bit The list captures a specific inflection point in deep learning, when the community realized that hand-crafting ResNets was tedious and began automating the automation. The breadth is genuinely useful: you can trace how NAS evolved from expensive RL-based methods (Zoph & Le, ICLR ‘17) toward more efficient techniques like weight sharing (ENAS) and differentiable search (DARTS).

Key highlights

• ~40 papers with direct PDF links, many with accompanying code repositories
• Covers three major NAS paradigms: reinforcement learning, evolutionary algorithms, and gradient-based/differentiable methods
• Includes hyper-parameter optimization papers (Hyperband, population-based training) as a distinct, related discipline
• Curated by Mark Dong, who also authored one of the listed papers (DPP-Net)
• CC0 license — effectively public domain

Caveats

• Appears largely unmaintained; newest paper is NeurIPS 2019, and the “awesome” ecosystem has since fragmented into more specialized lists
• No annotations or comparative analysis — just titles, authors, venues, and links
• Some code links are marked “not official,” which is helpful honesty but means quality varies

Verdict Worth bookmarking if you’re doing historical research or entering the NAS field and need a structured starting point. Skip it if you want modern 2023+ methods or critical evaluation of which techniques actually work in practice.