Machine learning portfolio tips 1. Good ideas come from ML sources that are a bit quirky. - NeurIPS from 1987 - 1997 - Stanford’s CS224n & CS231n projects - Twitter likes from ML outliers - ML Reddit’s WAYR - Kaggle Kernels - Top 15-40% papers on Arxiv Sanity
-
Show this thread
-
Replying to @EmilWallner
Alternatively: Chase your interests. Don't worry too much about making a portfolio or launching an idea. Tweet what you're doing. Merit rises on Twitter, at least for ML work. If you're worrying about getting upvotes on HN or Reddit, you might be wasting time.
1 reply 3 retweets 42 likes -
Replying to @theshawwn @EmilWallner
I think DeOldify would have been successful regardless of whether it got upvoted. Jonathan Fly built his following by tweeting out interesting results. Peter Baylies became known by making a good reverse encoder for StyleGAN then pointing it out when it was relevant.
1 reply 2 retweets 11 likes -
Replying to @theshawwn @EmilWallner
The common theme is that if you're doing good work, metrics follow. It's the work that matters, not the numbers. There might be survivorship bias in my analysis. What about ML researchers who do interesting work and never get noticed? But people tend to recognize good work.
1 reply 3 retweets 12 likes
This was exactly my thinking when doing DeOldify. I figured rather than focusing on gimmics, just do something I like and that would catch on. This was quite a deliberate decision. Easier said than done of course- the work is an absolute pain in the ass in practice lol
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.