chillee

19 ms·

151.
by chillee 6y ago
Latency of the typing is not the same thing as general UI responsiveness.
152.
by chillee 6y ago
A combination of low quality blogspam being posted there and the irritating paywall model.
153.
by chillee 6y ago
Yes you can. Cayley Hamilton theorem gives a linear relation between A^N and the terms less than that. So for your problem, you can calculate the number of paths from A->B of length k in k*N^2 time, you need O(N^3) to compute enough term
154.
by chillee 6y ago
> You can more easily generalize this solution to other arithmetic sequences. Fun trick, if you ever find yourself in a situation like this - instead of calculating the matrix by hand (a tedious chore), you can use an algorithm called Be
155.
by chillee 7y ago
For teams without as much funding, there's also a competition called the "First Tech Challenge". It's a smaller robot, which makes it far easier for a team to self-fund and get started. I started a team with 4 other frie
156.
by chillee 7y ago
No if you're just counting you can definitely do it faster than 2^n time. I believe you can do it in linear.
157.
by chillee 7y ago
Poor wording perhaps. I meant that there was no project that "officially" called itself a neural network library for Jax, not that Jax had "officially" chosen a neural network library. I'm talking, of course, about
158.
by chillee 7y ago
Certainly if you ignore those papers, you'd likely have no issue in practice - I suspect many of them are about more theoretical concerns. Perhaps I'll take a look at/post a list tomorrow. Either way, I believe that our origi
159.
by chillee 7y ago
I agree from a practitioner standpoint - but you were talking about research :) 1. Much of the scientific computing/stats community is stuck in the past. Many are still using Matlab! As opposed to the CS community, who are used to lear
160.
by chillee 7y ago
Trax is meant as a successor to Tensor2Tensor. From my understanding of what Tensor2Tensor was (and thus what Trax is): Basically, at it's core, it's a library of models, datasets, and utilities that are meant to be clear and work
161.
by chillee 7y ago
If you're talking about Jax, there's a couple different reasons to bother for research 1. Full numpy compatibility. 2. More efficient higher order gradients (because of forward mode auto diff). Naively it's asymptotic improve
162.
by chillee 7y ago
Super late reply, so it's likely you won't see this... (Too bad HN doesn't notify on replies). I'm not complaining about Flux in general, I'm talking about the specific example (the UNet) he brought up that he uses
163.
by chillee 7y ago
IMO, these kinds of functional abstractions look nice on paper, but are a pain in the ass to actually use. In practice, you'll want to print out things in between each layer, you'll want to log each layer's activations, you m
164.
by chillee 7y ago
I mean, this starts to sound like your complaint isn't that OpenAI is standardizing as a framework but that the framework is PyTorch :)
165.
by chillee 7y ago
OpenAI always had an "official" framework - it just used to be Tensorflow. That's why many of its public packages (baselines, spinningup, blocksparse, etc.) were built with Tensorflow. OpenAI has researchers, and it has peopl
166.
by chillee 7y ago
Actually, I believe the default hash functions have terrible hash quality - they're literally the identity function for integers. I do agree that unordered_map is extremely slow.
167.
by chillee 7y ago
I understand the sentiment, but O(n) vs O(1) and O(n^2) vs O(n log n) are huge jumps in complexity. Even with relatively small sizes like N=100, you're already starting with 2 orders of magnitude disadvantage. The example in this post
168.
by chillee 7y ago
Another piece of contextual information that's important is that interviewers can provide hints during an interview. For example, I was asked this question once: Given N points and 3 squares of side length L, what's the minimum L
169.
by chillee 7y ago
I think it's more related to how "normalized" it is. You don't really view wc as solving a problem you had - you view it as something that's always been there. It's worked for decades, and will likely continue
170.
by chillee 7y ago
Oh I read this as a joke... Like, don't encourage your SO to find someone better than yourself.
171.
by chillee 7y ago
Perhaps I'll provide a contrary view. I was also at Neurips, and I'm also a junior researcher. Yes, it was massive, and the poster sessions were crowded. However, I was still able to talk to everybody I wanted to talk to - often f
172.
by chillee 7y ago
Doesn't seem to have been a problem for React/PyTorch/GraphQL. Although perhaps I'd agree, actually. I'm not sure. It's at least not a huge deterrent.
173.
by chillee 7y ago
Funnily enough, the Tensor2tensor library he mentions that replaced the old deprecated layer is now also deprecated. https://github.com/tensorflow/tensor2tensor > It is now in maintenance mode — we keep it running a
174.
by chillee 7y ago
Uh... 60mph is ~27m/s. Are you saying that drivers need to accelerate to 60mph in half a minute in order to be comfortable?
175.
by chillee 7y ago
I'm very skeptical about numer.ai. I participated in their contests for quite some time (made a significant chunk of change), but I have huge doubts that any of our predictions are actually useful to them, for a multitude of reasons.
176.
by chillee 7y ago
You're not giving them enough credit. Oriol Vinyals (head of this project) is both one of the leading AI researchers + was highest ranked StarCraft player in Spain back in high school. Certainly the game is different, meta has changed,
177.
by chillee 7y ago
IMO, if you're buying a higher end laptop it's worth shelling out for the on site warranty/accident protection. They'll send a technician to your location (typically within the week) and replace whatever parts needed for
178.
by chillee 7y ago
I mean, the authors implemented it in pytorch: https://github.com/vsitzmann/scene-representation-networks Do you have any particular evidence that PyTorch is slow here?
179.
by chillee 7y ago
I wanted to write something about Julia/Swift, but I had a bit of a hard time writing a cohesive narrative about them and ran out of time. You can take a look at https://discourse.julialang.org/t/where-does-julia-p
180.
by chillee 7y ago
I talk about this in the "ablation" section/appendix. https://thegradient.pub/p/cef6dd26-f952-4265-a2bc-f8bfb9eb1e...

More ›