abhgh

31 ms·

301.
by abhgh 8y ago
Thanks!
302.
by abhgh 8y ago
But you won't be able to train it to the same accuracy. I'm not sure I agree with this bit in theory . A Neural Network is a stack of basis functions; and this stack can also be seen as a bunch of basis functions. And basis fun
303.
by abhgh 8y ago
If you need closer to a ELI5 version I recommend this - [1]. Disclaimer: written by me. [1] https://blog.statsbot.co/support-vector-machines-tutorial-c1...
304.
by abhgh 8y ago
Ha! Surprisingly this is not the first time I have seen someone describe this to be the boundary for logistic regression. (Btw agree with your other comment.)
305.
by abhgh 8y ago
One glimmer of hope is this is in Monterey - where the Moneterey Bay Aquarium is. I understand they have a few programs directed at conservation efforts. Let's hope they take this up.
306.
by abhgh 8y ago
If you like DBSCAN, I would recommend checking out OPTICS. It is not a clustering algorithm per se, but generates information from which clusterings for multiple settings of DBSCAN parameters can be "queried" cheaply. The DBSCAN a
307.
by abhgh 8y ago
I think one of the bigger reasons was something the article touches upon - image sharing. Internet connections can often be sketchy, and Whatsapps compressed image sharing often ran circles around its competitors. The other reason was the i
308.
by abhgh 8y ago
I didn't know of this, thank you!
309.
by abhgh 8y ago
The "production quality" to use a slightly commercial term, is top-notch.
310.
by abhgh 8y ago
Do you mean this - [1]? Yes, this was pretty cool! I esp liked how Mike, the person initiating the bet, initially was super-confident he'd win: It would be very easy to point out to him the impossibility of his task, but far more inte
311.
by abhgh 8y ago
Yeah, I realized that this would be potentially be confusing about what I meant, so I edited my comment.
312.
by abhgh 8y ago
Same here. Given Abel's Impossibility theorem [1], the headline got me very curious. Not that you couldn't solve for any case (the theorem talks about a general solution), but still interesting. [1] http://mathworld.wol
313.
by abhgh 8y ago
This resonates with me. For my long-running Python projects I usually have a file called small.py which has these, as you put it, micro-abstractions. Over time some stuff from the file gets into the larger projects ... and some stuff does
314.
by abhgh 8y ago
I had come across this library - [1] - a while ago that claims to be the open source equivalent of Vizier. Does anyone know how well justified the claim is? [1] https://github.com/tobegit3hub/advisor
315.
by abhgh 8y ago
FYI there is also a different "automl" suite of libraries maintained by Frank Hutters lab at the University of Freiburg - [1]. His group has been leading a lot of research in the area of hyperparameter tuning. [1] https:/&#x
316.
by abhgh 8y ago
The Markov chains we normally we hear of are first order Markov chains. The general notion being d-order chains. These are hard to compute (too many model parameters) but there has been quite some research on an interesting middle ground:
317.
by abhgh 8y ago
Interestingly, the author of D is on this comments page ... he did do D :-)
318.
by abhgh 8y ago
The hyperband (HB) paper you refer to was discussed (briefly) by Frank Hutter in this talk - [1]. He observes that the interpretation is not correct, and had the experiment run longer, you would've seen different behavior. In fact thei
319.
by abhgh 8y ago
One of their more interesting applications is in Bayesian Optimization (BO). You'd do BO optimization where evaluating the objective function is expensive. As against, lets say, where you might use gradient descent, where you navigate
320.
by abhgh 8y ago
Two comments have made me curious - do you think this could be locality specific (no, I do note you said "everywhere in Oakland", but still ....)? This was not a major road - I got off any major road, and walked quite a bit to thi
321.
by abhgh 8y ago
I am not sure why you find this odd. This could be locality specific - I didn't tell you where exactly my walk was. Nor did I imply if I walked the same distance anywhere in Oakland - anywhere in all 78 sq mi in it - I would invite the
322.
by abhgh 8y ago
Ha, that reminds me. I have a thing for trying out good cafes, and on my way to Berkeley, I decided to stop at one in Oakland. Till that point my ambit had been more or less limited to areas where s/w companies are + touristy spots. So
323.
by abhgh 8y ago
You are not doing away with the test/validation set. You are doing away with: 1. grid search - you would now only test for parameters that look promising (based on some criteria). Instead of covering the whole grid, you want to be
324.
by abhgh 8y ago
Today you can do much better - there is a whole active and evolving sub-field of solving hyperparameter tuning as an optimization problem. An ex is performing bayesian optimization using gaussian processes. I hope the research from the comm
325.
by abhgh 8y ago
Oops, sorry! To your point, I agree. They hype around the area has evolved much faster than the area itself.
326.
by abhgh 8y ago
What was not clear from your earlier comment, and is now, is that when you say AI you don't mean AI as is practiced by most of academia and the industry but the vision of Artificial General Intelligence (AGI). If so, yes, that's a
327.
by abhgh 8y ago
This is a very good point. While the final model might be a weekend of training, getting there is a lot more iterations/work.
328.
by abhgh 8y ago
This is an interesting comment - where would you draw the line between AI and applied statistics? A lot of AI which happens to be ML (not saying there is non-ML AI, just that a significant chunk of AI being practiced today is ML) also happe
329.
by abhgh 8y ago
No, you haven't driven stick shift until you have driven stick shift in Bangalore. Not to mention you'll have a new found appreciation for gears 1-3 (of the 5 or 6 possible levels you can go up to).
330.
by abhgh 8y ago
I think one of the challenges with using CD in the real world is the "unclean" input that need to be mapped to the various primitives and structures of CD, or stuff in the same vein that came after. Without a way to do that automa

More ›