Colin Raffel
colinraffel.com · Personal blog · on HN since 2016
Colin Raffel, a machine learning researcher and professor, writes about deep learning, JAX, music datasets, and research career experience.
What HN saysCommenters recommend his JAX tutorial as a great basics introduction and point to his Lakh MIDI Dataset as a directly downloadable music resource.
20HN people shared it
5linked it in comments
15stories submitted
153story points
4 months agolast shared
2016people per year2026
Most-shared posts
- 1A Call to Build Models Like We Build Open-Source Software5 people · 125 points · first shared 2021 · 26 comments on HN
- 2You Don't Know JAX5 people · 11 points · first shared 2019 · discussion on HN
- 3The Lakh MIDI Dataset v0.13 people · 2 points · first shared 2019 · discussion on HN
- 4My Year at Brain2 people · 5 points · first shared 2017 · discussion on HN
- 5Writing a Google AI Residency Cover Letter1 people · 3 points · first shared 2019 · discussion on HN
- 6GANs and Divergence Minimization1 people · 3 points · first shared 2018 · discussion on HN
- 7Reflecting on Two Years of Professorship1 people · 2 points · first shared 2022 · discussion on HN
- 8When will language models be good enough?1 people · 2 points · first shared 2026 · 1 comments on HN
- 9Tag:neural networks1 people · first shared 2016
- 10A roadmap towards machine intelligence1 people · first shared 2016
What HN people wrote when they linked it
wjessup · 2023-09-29
Great basics tutorial: https://colinraffel.com/blog/you-don-t-know-jax.html
LunarAurora · 2022-12-26
FYI MetaMIDI Dataset is the best (I found). you can request access here https://zenodo.org/record/5142664 if you haven’t already.
Lakh MIDI Dataset https://colinraffel.com/projects/lmd is smaller but directly downloadable
sapphireblue · 2016-07-31
Thank you! I found a great list of deep learning paper summaries that is mostly about learning algorithms from data and other cutting edge models: http://colinraffel.com/wiki/tag:neural_networks?do=showtag&tag=Neu… all kudos to the owner/author of this list !
>NNs that manipulate existing code? e.g. parse it, find design pattern, optimize
Also: you probably shouldn't need special architectures for such problems, these can be solved by seq2seq, NTM, RL-NTM and such, given a good dataset. Such tasks are still beyond current state of art though, it seems.
ArtWomb · 2023-03-02
NES APU sound synthesis chip is an ideal constraint for cross-domain learning with Transformers ;)
LakhNES: Improving multi-instrumental music generation with cross-domain pre-training
https://chrisdonahue.com/LakhNES/
The NES Music Database (5k+ scores)
https://github.com/chrisdonahue/nesmdb
The Lakh MIDI Dataset v0.1 (175k+ files)
https://colinraffel.com/projects/lmd/
visarga · 2021-12-12
Didn't say anything about algorithms improving faster than hardware. With better initialization and regularization, more efficient layers discovered by AutoML, sparsity, lower precision, or by directly combining models trained separately on diverse tasks instead of training sequentially we could have better models at the same hardware cost.
Even by improving the way people collaborate on model training we could make a huge leap. Instead of retraining from scratch we could reuse and update existing models. [1]
[1] A Call to Build Models Like We Build Open-Source Software https://colinraffel.com/blog/a-call-to-build-models-like-we-build-open…