abhgh

27 ms·

241.
by abhgh 6y ago
Same here ... for technical textbooks that I really need to focus on, dead tree is better. Esp if the content is math-y, scribbling notes, underlining etc seem to amplify learning.I bought a printer just for printing out such content so I c
242.
by abhgh 6y ago
Your link [2] doesn't work. If you were linking to Incropera-Dewitt its a great book!
243.
by abhgh 6y ago
I have a vintage fx-82 calculator [1] that still works. My father bought it when he started working after his engineering, which was passed on to me for my engineering, which was then passed on to my brother for his engineering - which ende
244.
by abhgh 6y ago
Traditional Automatic Speech Recognition (ASR) systems do this, and this component is known as a Language Model (LM). Typically you would use/train a LM for your domain or specifically for your dataset.
245.
by abhgh 6y ago
This was part of an educational kit I played with, as a kid: Chip-Chap [1]. Was too young to understand the "how", but getting things to work by following the manual was an absolute delight! [1] http://hopesprogramkits.
246.
by abhgh 6y ago
There has been some work in grounding NN conversations by either giving them factual memory/knowledge base - [1], [2],[3]. Personally I would love to see more work in this area where you have a robust language generator guided by a (1)
247.
by abhgh 6y ago
Its a very handy utility. Sometimes when I am lazy to write proper multiprocessing support in my codes, I get some work started anyway by (1) quickly providing a basic CLI version of my code (2) throwing in a bunch of code-invocation comman
248.
by abhgh 6y ago
Not very commonly discussed, but great topic to be aware of. One particularly powerful use of this is in learning models with interaction terms: more powerful than a linear model but still interpretable, [1]. [1] R package https:/&#x2
249.
by abhgh 6y ago
Oh I remember this! Was hilarious!
250.
by abhgh 6y ago
I would add Bayesian Optimization as a very strong member of family of derivative free algorithms. It assumes a black-box function, which also can be noisy (not all derivative-free algorithms can work well with noise). One of my favorite ca
251.
by abhgh 7y ago
Convenient to use and very helpful for technical illustrations. For ex, I created the gifs here [1] using LICEcap - showing a class boundary in 3D space. Have been using it at work for a couple of years now. [1] https://blog.stat
252.
by abhgh 7y ago
I haven't had time to write a blog post about this. The paper linked to in the first comment - which is [1] - present the first cut of our ideas. We have made two additional extensions beyond that, we will be submitting them to journal
253.
by abhgh 7y ago
I am! And the lottery ticket (LT) hypothesis is something I have thought about quite a bit. I believe the LT hypothesis and our work are related in a subtle way: in the former we build out the larger structure first and then prune, while i
254.
by abhgh 7y ago
Very research-y, but I am working on a pet idea of mine called "compact models" in Machine Learning. Essentially you take a model - say a decision tree of depth 20 - and "compact" it to a decision tree of size 10, withou
255.
by abhgh 7y ago
I really hope so. Including SIFT in OpenCV should be easy since they did have an implementation earlier, which they removed due to the licensing.
256.
by abhgh 7y ago
Recommendation seconded - for people with a fascination for secret doors!
257.
by abhgh 7y ago
The first time Google contacted me, I didn't make it (and was not happy with their process but thats another story). Thats when they let me know that they interview people on various things first and then decide the role. When they rea
258.
by abhgh 7y ago
Definitely - I went through some of his lectures after I had read up on Linear Algebra as part of a course (from [1]) and had used it in a bunch of applications, but I still found the lectures worth going through if only because his fantast
259.
by abhgh 7y ago
> If the goal... I disagree here. This is not always the goal of every research student. Some set the bar at solving a problem, and not going through a fixed number of mechanical steps of learning how to do research. I think otherwise to
260.
by abhgh 7y ago
Not a Santa Cruz resident, but I have occasionally worked out of The Abbey, and see students there.
261.
by abhgh 7y ago
Thats what I thought too.
262.
by abhgh 7y ago
Seconding TUI. Pretty solid. In fact if you need a better debugger than that I see no reason as to not jump to a good IDE ... no other middle ground required.
263.
by abhgh 7y ago
I found them okay-ish. In fact, I felt the cooking was not uniform because it felt more "bready" towards the periphery (yes its supposed to be like that, but this was more than standard pizzas). NB: things may have improved, I had
264.
by abhgh 7y ago
Was going to just say this!
265.
by abhgh 7y ago
Amazon accidentally shipped me a book a few months ago, by getting one digit in the address wrong. I pinged Amazon to return it - but I got the same response, "why don't you keep it?". I said I didn't want to, and this w
266.
by abhgh 7y ago
Brings back memories. I think implementing an AI for tic-tac-toe was one of my first complete implementations of a RL algorithm; circa 2008, I think :)
267.
by abhgh 7y ago
I agree that MMDS is probably not a good introduction for ML- pretty idiosyncratic choice of topics for that. If you are fine with lectures, I always recommend Abu Mostafas lectures [1]; I think they have an extremely good balance of intuit
268.
by abhgh 7y ago
Yes this is a good book, esp. for an overview. At one time, I used to organize weekly discussion sessions at work based on sections in the book :-)
269.
by abhgh 7y ago
Same with their Roku app - I quickly jump from one thumbnail to the next to avoid triggering the autoplay.
270.
by abhgh 7y ago
There is a downside to this - there are no clear work hours boundaries. My father who is a MechE is annoyed that I am often running to my laptop to answer mails, or to code etc beyond a fixed 8-9 hr in a day. Btw my bachelors was in Mech to

More ›