Grad student at The University of Tokyo. Curious about how we as humans make sense of what we see!
Japan
Joined August 2010
- Tweets230
- Following110
- Followers197
- Likes374
Jayani Withanawasam. retweeted
I'm happy to announce major changes to Tokyo Data Science @TokyoDS:
(1) Fair Play Tuition makes it affordable to *anyone* acalonia.com/fair-play
(2) You can join even for a course or two
(3) Name changed to Acalonia @acalonia_x since it's global
Pls retweet for visibility🙏
Jayani Withanawasam. retweeted
Thanks to the JAX/Flax community week, I had a chance to fine-tune a CLIP model on a radiology dataset (ROCO) on cloud TPU-v3-8. It was a quite fun learning experience.
☞Github repo: github.com/Kaushalya/medclip
☞@streamlit demo on @huggingface 🤗 Spaces: huggingface.co/spaces/kausha…
Jayani Withanawasam. retweeted
"The first step of any project is to grossly underestimate its complexity and difficulty."
- Nicoll Hunt @nicollhunt
Jayani Withanawasam. retweeted
When we get a new idea, our impulse is to look for signs that could validate it. But the most effective way to make real progress is to look for the simplest way to prove your idea wrong.
Jayani Withanawasam. retweeted
Clean Pytorch implementations of ML techniques with explanations next to the code. This is a format that makes understanding papers way easier. @__MLT__
Annotated PyTorch implementation/tutorial of Feedback Transformer by Fan et al.
lab-ml.com/labml_nn/transfor…
Github: github.com/lab-ml/nn
@facebookai @EXGRV @armandjoulin
Jayani Withanawasam. retweeted
Woke up to a rejection letter which sucked. But at least I got to add a sticker to the page.
Jayani Withanawasam. retweeted
I passed this image through a pretrained ResNet 101 with @pytorch and it predicts:
Tiger shark 23%
Hammerhead 21%
Great white shark 16%
Gar, garfish 11%
Sturgeon 3%
My conclusion : shape > [ texture + mountain context]
=> reassuring in some way.
(Image from @SolTight)
This Post is from an account that no longer exists. Learn more
Jayani Withanawasam. retweeted
We're bringing you the 2nd episode of the Stanford MLSys Seminar tomorrow. @matei_zaharia will talk about lessons from @databricks in building and deploying @MLflow.
Tune in at 3pm PT Th at youtube.com/watch?v=nCQ9WqXP… (and join our mailing list at groups.google.com/forum/#!fo…)!
Jayani Withanawasam. retweeted
I don't mind electronic gadgets, but I draw the line at connected ones. Those tend to have more power over you than you have over them
Jayani Withanawasam. retweeted
If you're not happy single, you won't be happy in a relationship. Happiness comes from buying new stationery.
Jayani Withanawasam. retweeted
Most useful tip for doing presentations over Zoom. Share only a portion of the screen. Rest of the screen can be used for anything - timer, speaker notes etc.
youtube.com/watch?v=gx1Ysf_e…
Jayani Withanawasam. retweeted
Our paper "MetAL: Active Semi-Supervised Learning on Graphs via Meta Learning" [arxiv.org/abs/2007.11230] has been accepted for ACML 2020. 🤘🏽
Code: github.com/Kaushalya/metal #AcademicTwitter
Jayani Withanawasam. retweeted
Crowdsourced generative art gallery 👇🏽
This is art created and described by 66 anonymous individuals on Amazon Mechanical Turk. They used a “Create Your Own” tool from cc.gatech.edu/~parikh/art.ht… to make these. I'll post one ~every week.
#generativeart #creativecoding #crowdsourced
Jayani Withanawasam. retweeted
In support of #free_science, 10s of publicly available #ECCV2020 videos are available at: crossminds.ai/category/eccv%…
Any attempt of blocking science behind a virtual platform is unlikely to success as authors aim for maximise audience and impact of their work.
Jayani Withanawasam. retweeted
Many ML practitioners have no idea where to start getting a foothold in learning theory. I certainly didn't. "Understanding Machine Learning" by Shai Shalev-Shwartz & Shai Ben-David is an absolute treasure. Beautiful hardcover & free download @ cs.huji.ac.il/~shais/Underst…
Jayani Withanawasam. retweeted
Thanks to everyone that replied to my question on what the AI community should work on. I think if we come together as a community we can make better progress. I wrote about this in The Batch today (reposted here). Would love to hear your thoughts!
Jayani Withanawasam. retweeted
"Counterintuitive Aspects of High-Dimensional Distributions" by @fabinger is online! Unusual, surprising, fun, and a lot of aha! moments – enjoyed it a lot, thank you Michal! :) #math #atitsbest
🍉 youtube.com/watch?v=RwDWCB8D…
Jayani Withanawasam. retweeted
Good benchmarks are important for progress in ML, but personally I’m more interested in research that explore new methods capable of doing things that existing SOTA methods simply fail at doing.
I also think pursuing such research is easier and more enjoyable than beating SOTA.
Jayani Withanawasam. retweeted
At Tokyo Data Science @TokyoDS, our statistics series covers topics one would learn in their first year of Harvard econ PhD. I can write grad school recommendations for talented students. The tuition is free for low income people in developing countries: xitter.gq/fabinger/status/128478…
We have a good number of full scholarships at Tokyo Data Science @TokyoDS for a remote group starting in July. They are e.g. for low-income people in developing countries or those with social impact. First: Stats and Neural Nets. DM me if interested. Please retweet for visibility