abhgh

10 ms·

91.
by abhgh 2y ago
Since we are discussing Mondrians in the context of trees, I thought I would mention "Mondrian Forest" [1]. Interesting but relatively unknown algorithm to construct random forests over online data. [1] https://arxiv.or
92.
by abhgh 2y ago
If you don't mind a tangential question, what are some good places to watch condors in CA? We watched them at the Pinnacles National Park and loved the experience.
93.
by abhgh 2y ago
If you're in the bay area he has an upcoming event tomorrow, at Kepler books [1], where he would be talking about AI in education and his new book Brave New Words [2]. [1] https://www.keplers.org/upcoming-events-inter
94.
by abhgh 2y ago
To add to a sibling comment, if you're interested in learning a bit about the both the Gaussian (as in a density estimator like Gassian Mixture Models, aka GMMs) vs Gaussian Processes (GP), I have some write-ups here: [1] and [2]. [1]
95.
by abhgh 2y ago
This is sad news indeed. The Simons Institute [1] in the UC Berkeley campus has had a positive impact in my life in terms of the many high quality talks (both in terms of content and recording quality) that they continue to put up on YouT
96.
by abhgh 2y ago
I wasn't aware of this till a friend told me: the Simons foundation partly also funds the Perimeter institute [1]. [1] https://perimeterinstitute.ca/news/new-simons-foundation-sup...
97.
by abhgh 2y ago
You're welcome! There is a talk too [1], which is how I had learned about the paper. [1] https://www.youtube.com/watch?v=sD-uDZ8zXkc
98.
by abhgh 2y ago
Yes, also TriMap does quite well. PacMAP is faster though. This paper [1] (linked at the GitHub repo you have mentioned) goes into a fantastic amount of detail in comparing UMAP, t-sne, PacMAP and TriMAP. [1] https://jmlr.org&#x2
99.
by abhgh 2y ago
I think a common problem is that these techniques get repurposed to solve problems that they weren't meant to. I have seen multiple people fall too often into the trap of using these visualizations to guess whether a dataset may be cla
100.
by abhgh 2y ago
Sorry to hear that. My experiences haven't been very different. I really can't tell if the current review process is the least bad among alternatives or is there something better (if so, what is it?).
101.
by abhgh 2y ago
Yes arxiv is a good first source too. I mentioned conferences as a way to get exposed to diversity, but not necessarily (sadly) merit. It has been my experience as an author and reviewer both that review quality has plummeted over the years
102.
by abhgh 2y ago
Yes seconding this. If you want a broad view of ML IMHO the best places to look at are conference proceedings. The typical review process is imperfect so that still doesn't show you all the interesting work out there (which you mention
103.
by abhgh 2y ago
Thanks for mentioning that. It's a problem I occasionally need to solve and based on your comment, I found this helpful write up by Nick Higham [1]. [1] https://nhigham.com/2021/01/26/what-is-the-nearest-
104.
by abhgh 2y ago
Another issue I've personally faced is debugging - although I am saying this from my experience from more than a yr ago, and maybe things are better now. I have used it mostly for optimization and the error messages aren't helpful
105.
by abhgh 2y ago
Sorry if I didn't make that clear: I wasn't saying they are methodically equivalent (although there are similarities, e.g., the Dirichlet Process is a distribution over multinomials, which would be one way to formulate a clusterin
106.
by abhgh 2y ago
A fairly interesting topic that I think remains important to this day. Some notes: * The article misses out on HDBSCAN [1], which is a relatively new technique, and quite effective. * With the arrival of embeddings, esp. sentence embeddings
107.
by abhgh 2y ago
They are like "continuous" databases. See slides 4-5 here [1] - this is from a talk I had given a while ago. [1] https://drive.google.com/file/d/12uHo9QIfS-jBpVTs3lmQ3BEpxhD...
108.
Bayesian Optimization, Part 1: Key Ideas, Gaussian Processes(blog.quipu-strands.com)
3 points by abhgh 2y ago|0 comments
109.
Fun with GMMs(blog.quipu-strands.com)
2 points by abhgh 3y ago|0 comments
110.
by abhgh 3y ago
"nonparametric" is somewhat of a (confusing!) misnomer in that it doesn't mean no parameters, but lots of them where the # of parameters grows with # of instances [1]. In all of these cases the models have some general parame
111.
by abhgh 3y ago
Yes I did, thank you!
112.
by abhgh 3y ago
It becomes valuable if it either compensates for the lack of hyperparam tuning in all or a broad class of models, or the result is precisely presented to say something like "better than random if the model is a linear svm with C=1&quot
113.
by abhgh 3y ago
Thanks for the good discussion! * Reg. model selection/calibration: I acknowledge that in interactive setups cross-validation (CV) is not practical, but that is a particular kind of model selection. And faster alternatives exist for sp
114.
by abhgh 3y ago
Appreciate the reply! We have 2 binary datasets in there [1], but of course, those results are not separately presented - you see the averages. The choice of datasets here is based on what other papers report, to have comparable results. In
115.
by abhgh 3y ago
I second the point about learning a classifier over them - it is practically quite useful. But I'd caution against believing Active Learning (AL) would be unequivocally be useful; most positive performances seem to arise in very speci
116.
by abhgh 3y ago
Not for DL models, but I was exploring doing this in a specific setting (I've posted about this earlier). For small-sized models (for some reasonable definition of size, e.g., depth of a decision tree, or # trees in a gradient boosting
117.
by abhgh 3y ago
Variable Discovery. Yes that isn't a thing; its a term I came up with, and I sincerely hope no one ever has to encounter such a situation. So, someone wrote a complex data processing pipeline, that read from some tables in a DB, proces
118.
by abhgh 3y ago
I have found learning JAX to be rewarding. Of course this depends on the work you do - but if you want to occasionally code out problems which involve some kind of an optimization of a differentiable quantity (without you having to work out
119.
by abhgh 3y ago
I had worked with it for a while (when I was at Adobe, a lifetime ago), and I thought it was a cool piece of technology. You could build up complex good-looking UIs fairly quickly. Was sad to see active development stop.
120.
by abhgh 3y ago
Seconding the book recommendation. The author is Jeff Deutsch who is the director of the Seminary co-op bookstores, Chicago. These also have been around for a while. I have been to the one in the Univ. of Chicago campus, and would recommend

More ›