mesuvash

5 ms·

1.
by mesuvash 6mo ago
IIUC, The paper's notation S^(d-1) means the unit sphere in R^d (e.g., the familiar unit circle is S^1 living in R^2). So, i think, x in the algorithm is already a unit vector. Reference: Section 2:Preliminaries ... We use the notati
2.
by mesuvash 6mo ago
Fair point. I've updated the animation to address this. The grid now uses the correct non-uniform centroids (optimal for the arcsine distribution in 2D), so you'll see grid lines cluster near the edges where unit-circle coordinate
3.
by mesuvash 6mo ago
Yes, this is important in high dimension. But sadly, very hard to visualize. In 2d it looks like unnecessary.
4.
by mesuvash 6mo ago
That's actually correct and intentional. TurboQuant applies the same rotation matrix to every vector. The key insight is that any unit vector, when multiplied by a random orthogonal matrix, produces coordinates with a known distributio
5.
by mesuvash 6mo ago
Yes. Great catch. I simplified the grid just for visualization purpose. I've updated the visualization. The grid is actually not uniformly spaced. Each coordinate is quantized independently using optimal centroids for the known coordin
6.
by mesuvash 6mo ago
Author here. Sorry still working on refining the post. Will share once the post is ready.
7.
by mesuvash 6mo ago
TurboQuant explained with an easy to understand (no-math) animation https://mesuvash.github.io/blog/2026/turboquant-interactive/
8.
by mesuvash 7mo ago
I am glad you liked it :) You might like this https://mesuvash.github.io/blog/2026/rl_for_llm/ as well :)
9.
DeepSeek's Dualpath Paper explained with animations(mesuvash.github.io)
2 points by mesuvash 7mo ago|0 comments
10.
Intuitive Intro to Reinforcement Learning for LLMs(mesuvash.github.io)
3 points by mesuvash 7mo ago|0 comments
11.
An Intuitive Introduction to PPO and GRPO(mesuvash.github.io)
5 points by mesuvash 7mo ago|2 comments
12.
by mesuvash 8y ago
Thanks for the pointers. From my personal experience, Auto-encoders are amazing for dense input (images, audio etc), more specifically, when the input feature space is not large. However, in many real-world problems such as recommendation,
13.
by mesuvash 8y ago
>>I think the hashes could take some work. Any suggestions or thing that are not clear? Thanks for your feedback. I shall update the post accordingly.
14.
by mesuvash 13y ago
Nothing but #respect. It's hard to see people who give up fortune for what they consider right thing(atleast from his perspective).
15.
Seaborn: Amazing plots with matplotlib(stanford.edu)
2 points by mesuvash 13y ago|0 comments
16.
by mesuvash 15y ago
Thanks for the info :)
17.
by mesuvash 15y ago
Btw, If someone wants to drop out from the course. How can he/she do so ?
18.
by mesuvash 15y ago
Awesome. MITx platform is superior than any other online learning platform i have ever seen. Very well done. Congrats.
19.
Show HN :Interesting videos on the web Lolzwow.com
1 points by mesuvash 15y ago|0 comments
20.
Show HN: Lolzwow.com Interesting videos on the web
1 points by mesuvash 15y ago|1 comments
21.
by mesuvash 15y ago
videolectures.net videos are generally about technical topic. Mostly related to Machine learning these days.
22.
by mesuvash 15y ago
If you are looking for inspirational videos or Interesting videos you can have a look at http://www.lolzwow.com
23.
Why "Compilers: Principles, Techniques, and Tools" is must read for CS students.
6 points by mesuvash 15y ago|2 comments
24.
by mesuvash 15y ago
Well i think you need to adjust your routine. I think both startup and relationship are equally important and you have to manage accordingly. Relationship stands on top of understanding and i think you both should understand each other. If
25.
by mesuvash 15y ago
Yeah, i am in verge of adjusting my expectations.
26.
by mesuvash 15y ago
Watch these videos http://bit.ly/ehfTc3 http://www.youtube.com/watch?v=UF8uR6Z6KLc http://bit.ly/eNZPda I do it whenever i lack motivation. It helps me a lot, I hope it will help you in same way. And you if you really think you want t
27.
by mesuvash 15y ago
Nice work.
28.
by mesuvash 15y ago
I will suggest you to start with the one you are most comfortable with. Both has strong community. Being a pythoneer, I can assure that python community and IRC is very has lots of people who are really very helpful. I dont know much abou
29.
by mesuvash 15y ago
Yeah sometimes i feel TechCrunch is biased. I am not comfortable with people who write news on behalf of some company. We expect them to write the real news rather than supporting some and reprimanding others.