abhgh

28 ms·

1. 271.▲
   by abhgh 7y ago

   You're right, thats what the term seems to stand for (I had Googled too, when I was thinking of names). "Compact models" is something I am considering calling this work - its by no means standard :)
2. 272.▲
   by abhgh 7y ago

   I am working on something I am calling "compact models" (as part of a PhD): techniques to pack more information into Machine Learning models when their size is constrained in some way. I have put up some of our work here [1] - it
3. 273.▲
   by abhgh 7y ago

   The intermediate R layer is often a well thought out abstraction - hence it makes a lot of sense to re-use it. At different points in my career I have written skunkworks wrappers for rpart (decision tree learner) and glmnet (generalized
4. 274.▲
   by abhgh 7y ago

   Yeah its a weird article ... the author accuses TeX of being a typesetting language, which is what TeX advertises itself as anyway.
5. 275.▲
   by abhgh 7y ago

   It happened this year and it was packed. So much so that when I reached 10 min late, there were so many people (standing) that I couldn't actually see Knuth even though I was technically inside the Nvidia auditorium, Stanford. Could
6. 276.▲
   by abhgh 7y ago

   Oh that's surprising. I uploaded content generated by overleaf (they provide a arXiv compatible dump of your project) a month or so ago and it worked smoothly (after uploading to arXiv I deleted the bib file of course).
7. 277.▲
   by abhgh 7y ago

   Just delete the .bib file. arXiv only needs the .bbl file. What is confusing is even if you have both the .bib and .bbl, arXiv doesn't ignore the former, it raises an error anyway. Bad error reporting. Another tip if you're us
8. 278.▲
   by abhgh 7y ago

   I have given away a couple of Klein bottles to collaborators I have really enjoyed working with, as a token of appreciation. He sends emails that show him packaging, and a bunch of eccentric paperwork accompanies the bottle, which are a del
9. 279.▲
   by abhgh 7y ago

   I have been working on AI/ML since 2007, and this is what I had read too (repeatedly). That Go is incredibly hard to conquer for a computer, and it will be a long while. So when this actually happened, I was shocked/surprised.
10. 280.▲
    by abhgh 7y ago

    Thanks for adding that to the discussion. I'd like to point out a couple of things: (1) that adding features can create problems is well known among good ML practitioners (I daresay, esp. to those who have a fair amount of exposure to
11. 281.▲
    by abhgh 7y ago

    I wrote a starter article quite a while ago about using Naive Bayes for analyzing clickstream data if anyone is interested: https://tinyurl.com/ulmam85 Recommend downloading the PDF for better rendering of math symbols.
12. 282.▲
    by abhgh 7y ago

    This was my reason for foraying into Tcl too. Till then I had used Java's AWT and Swing; but prototyping with Tcl/Tk was way faster and intuitive. Mid-2000s. I ended up liking the language - idiosyncratic as it was. Partly the r
13. 283.▲
    by abhgh 7y ago

    This is interesting! So its not just about grammatical patterns, but other stuff that might be parsed as named entities.
14. 284.▲
    by abhgh 7y ago

    Here's a variation I've seen: bad guy finds jewelry (necklace/ring) near you, making sure you've spotted it too. Then you get in an argument for ownership. He suggests splitting the cost - he would leave if you pay him,
15. 285.▲
    by abhgh 7y ago

    LightGBM occupies a sweet spot between speed and accuracy, and is a library I've grown to love. Nowadays, this is my primary choice for quick impactful results. NB: if your data has categorical features, you might easily beat xgboost i
16. 286.▲
    by abhgh 7y ago

    If you want to begin with what your outcomes could look like if you use TeX and friends, as suggested by jkepler, see here - [1]. [1] https://tex.stackexchange.com/questions/1319/showcase-of-bea...
17. 287.▲
    by abhgh 7y ago

    Thanks!
18. 288.▲
    by abhgh 7y ago

    In the context of attention, there is a very interesting recent paper that warns against conflating attention and token importance - "Is Attention Interpretable?" [1]. This is an accepted paper in ACL-2019: [1] https://
19. 289.▲
    by abhgh 7y ago

    I was going to mention XLNet before I saw your comment. Also, a recent piece of interesting work [1] shows that with the right control parameters, you could still use gated RNNs, like LSTMs, for pretty good language modeling. [1] http:&#x2
20. 290.▲
    by abhgh 8y ago

    I quite like his presentation format for his online lectures (example - [1]); 3 panels - (1) video of him explaining/board work (2) notes (3) slides, all sync-ed. When he changes topic the slide and notes panel also move forward. And I
21. 291.▲
    by abhgh 8y ago

    "Regression" shouldn't be a separate category of algorithms given that the other categories are learning algorithms like ANNs, SVMs etc which are also used for regression. Also, it mentions Bayesian algorithms are fast - that
22. 292.▲
    by abhgh 8y ago

    I tried taking the train from Santa Barbara to San Jose - I have heard a lot about the picturesque route - but the train was delayed at arrival by a few hr, and since I had a prior engagement, I'd to rent a car and drive to SJ.
23. 293.▲
    by abhgh 8y ago

    I think this is one of those things that depend on your actual workflow, content etc. I see your point; for me this hasn't been a problem.
24. 294.▲
    by abhgh 8y ago

    I'd also add that for figures Inkscape is invaluable [1]. Save as svg once, and export it as whatever later. I typically export it to PDF (from within Inkscape) for pdflatex. While its typically indispensable for schematics, I often se
25. 295.▲
    by abhgh 8y ago

    I recommend Dell's Alienware (AW) laptop if you need a machine with good specs that can reliably run Linux. I have been a happy Ubuntu user on AW for a while. Great options for hardware, built like a tank, and can run simulations/
26. 296.▲
    by abhgh 8y ago

    I hope it grows in popularity if only because its an interesting take on learning. I did a course on RL in 2007 and our textbook was the 1st edition of this book - back then, it was perceived to be a very niche area and a lot of ML practiti
27. 297.▲
    by abhgh 8y ago

    +1 If you're a bibliophile I do recommend stopping by their store in LV. I discovered it by accident but its quite a treat.
28. 298.▲
    by abhgh 8y ago

    While this is mostly true there is has been some research to use non-GP models for Bayesian Optimization or the general Sequential Model Based Optimization (SMBO) problem. Some examples: * SMAC (using Random forests) - [1] * Using Deep Neur
29. 299.▲
    by abhgh 8y ago

    I think you're exactly right about the modularity aspect of DL; in fact I made a similar comment on this page, albeit speaking in terms of basis functions. I have a minor nitpick regarding this point you make: not that anything is lik
30. 300.▲
    by abhgh 8y ago

    Thanks! Yes its a pretty good demo, it should be more popular IMO.