Great answer. It's just a data point. There is no evidence that P(office manager, biomedical engineering) != P(office manager) * P(biomedical engineering).
I was a Production Manager at a Medical Device company here in the Bay area previously for almost 3 years. Wanted a change of scenery so I started helping Garry out.
From there, I started helping out with events and tech/ops at YC and joined the team to help out internally, officially (Tara and co have been doing so much already)! All I've got to say is I came for the experience, and stayed for the community.
I don't know about the rest of their stack, but their Machine Learning team is pretty good. They were using PMML the last time I spoke to them. And a little bit of Scalding as well. Preventing fraud at scale is quite a challenging exercise.
Actually approachable math - senior undergrad math major or first year grad can easily understand what's going on. ADMM is actually very old primal dual setup solved via Lagrangian, goes back to 1980s I think.
Good reference - https://web.stanford.edu/~boyd/papers/pdf/admm_slides.pdf
The article itself refers to Boyd's Convex Optimization textbook, which is sort of a rite of passage for anybody doing anything math-related at Stanford these days - even their MS in Management course requires a semester of convex optimization.
Stupid nature paywall honestly makes me furious. Why doesn't altman or some other yc partner throw some money at nature & get all their articles up on the web for free ? /endrant
> Stupid nature paywall honestly makes me furious. Why doesn't altman or some other yc partner throw some money at nature & get all their articles up on the web for free ? /endrant
One of the strange things about the software business is that software is an iterative process where the rate of iteration can go from linear to polynomial to exponential very fast. Another strange thing is that if you are a software guy, the whole world is your customer. The third strange thing is the constant reinvention of old wisdom, repackaged in new frameworks, languages, platforms.
These three strange things then combine in weird ways to take on a life of their own.
Had BCC been written in a prior era, it would have been some MFC code that ran only on windows 95 boxes, shipped on a dozen floppies, and purchased by atmost a few dozen people who bought a printed copy of Dr.Dobbs and looked at an ad for BCC and said, Hey, I need to make my own bingo cards.
But because he had a web app and priced it right and iterated upon feedback, the whole world came knocking on his door.
To be fair, when he first started writing about BCC it wasn't a web app. The web app came as part of the iterative process (not that this detracts from your argument).
Luckily, I was at mlconf last week where Dr.Anandkumar spoke - they called her the "tensor lady" :) She's using tensors in machine learning for a bunch of things -
Latent Variable Models: Training LVM's using local search methods like EM, gradient descent, variational bayes etc. have a bunch of problems - they get stuck on local minima, the algorithms are hard to parallelize with poor convergence. In these cases, tensors yield guaranteed learning using embarassingly parallel algorithms, so faster convergence & can be run on Spark.
Also saw a demo on training 2-layer nets for GMM using tensors, and they learnt the weights rather fast. So using tensors in deep learning shows promise, though the techniques are in their infancy.
One of the challenges the professor mentioned was the availability of open source libraries to do tensor decomposition, which the above methods require.
This isn't quite right. Moment methods (that rely on tensor decompositions) have a few problems:
(i) they have convergence bounds, but in practice need more data than we have available
(ii) they don't do as well as EM usually, but using them to initialize parameters for EM sometimes does better than EM with random initialization schemes
(iii) it turns out variational methods can also be embarrassingly parallelized without losing much accuracy in practice
(iv) right now moment methods don't work for arbitrary graphical models
I believe while you are right in general, she is looking at a class of problems for which tensors handily triumph other methods. You might be interested in these papers -
Is this a particular style of writing ? He takes 3 sentences, makes 1 paragraph. Then repeats with the next 3, and the next 3 etc. The paragraphs are completely disjoint - so he mentions almonds in a para, then Romney in another, then Glenn Beck, then something about Yosemite....its like a whirlwind tour of the drought landscape with no central thesis or conclusion. No flow, just lots of disparate facts. What is his point ? How exactly does one go from one freak drought to the end of California, which has what - the fourth or fifth largest GDP on the whole planet ( if it were a country ) ?
I agree. He's strung together a bunch of anecdotes and currently-voguish facts, without doing his homework and calling our attention to the most important effects. He has no real POV about how this plays out, he just seesaws between optimism about creative solutions and worry.
I'm a NYT subscriber, so I'm pre-disposed to their viewpoints and style, but this column is really junky.
I wonder if articles are being written so as not to scare off hyperlinked visitors. People are perhaps turned off by large paragraphs and want to be able to zoom in on some keywords which can be skimmed around.
Interesting, thanks, but those are 2013 numbers. The article claims California grew/is growing faster than Texas which isn't the case in your linked report, so I wonder where one might find more recent data.
This is a stunning book. 15 years ago, when I first came to this country from India, a classmate gave me a copy of this book. I guess he wanted me to understand America or Americans or whatever. It definitely affected me at some deep level. I still have my copy, and I read it everytime something upsets me about my job & I get those "maybe I should just switch to something else..." feeling.
Its not "by Study Terkel" because he didn't really write this book....I mean, its his book, but he basically ran into all these chaps, and let them do all the talking. And they talk and talk and talk. He prompts them from time to time, but mostly, its a book by these people. And their stories and lives are super compelling. I often used to wonder why people do dead-end jobs like parking cars in a building, or store clerk etc. Here, those people tell you why. There is this hooker who tells you how she got into the business, how she turns tricks, dealing with cops, offering freebies to pimps,the whole thing is simultaneously amazing and sordid and just puts everything into perspective. It taught me that America is a hard country. Its not some first world paradise. America is like, seriously fucked up. I mean, all these people, they were here long before I got here, and they have friends & families & homes & connections, and yet their lives are so hard and messed up and its not all whining & complaining, but I did get the impression things are very hard out here for the blue collar guy. Ofcourse had Turkel gone to Sand Hill road & bumped into a hundred VCs, that would be a very different book. It a definite must-read, especially if you are an immigrant.
