textminer

8 ms·

61.
by textminer 13y ago
I've just timed it out for about 45 minutes in the car, and an hour and fifteen for the BART/shuttle combo. I also quite like podcasts, so am all right with having just my ears free while driving.
62.
by textminer 13y ago
About to start a daily commute from Berkeley to Menlo Park. Am optimistic about 880-South in the morning, and driving seems far faster and more peaceful than the BART -> Shuttle method. But I do mourn the loss of productive time on a da
63.
by textminer 13y ago
My suggestion is to avoid telling a potential new employer what you make now. They'd then work hard to cap you at current + 30% instead of what a comparably-skilled peer might make.
64.
by textminer 13y ago
Random Forests are nice in having few parameters-- number of trees, number of restricted features to sample on at each decision node. It's the variations on the standard RF that have interested me. Rotation Forests, where features are parti
65.
by textminer 13y ago
I work in machine learning. I regularly get jazzed when I figured out a linear solution to a problem (e.g., using a min heap to grab top k elements from a similarity computation in a recommendation system instead of first sorting), and will
66.
by textminer 13y ago
Learn and be well: http://www.iro.umontreal.ca/~bengioy/papers/ftml_book.pdf Explains and motivates the use of autoencoders (and sparsity constraints) and restricted Boltzmann machines, too.
67.
by textminer 13y ago
Am I crazy, or is the short-run benefit more for unsupervised feature learning more than classification accuracy? If I recall from my brief reading on the subject, kernel-aided SVMs and Random Forests are basically equivalent to a three-lay
68.
by textminer 13y ago
For the long-term career-conscious and money-hungry, it almost makes you want to plan a trajectory that doesn't plateau at "senior engineer". That seems like a shame/misuse of resources, doesn't it?
69.
by textminer 13y ago
In this vein, it's been hard to fully appreciate San Francisco. I can "see" the lovely homes in Duboce Triangle and the Castro and in Hayes Valley, and I can "enjoy" the way the earth bends around you on a North Beach rooftop. But being in
70.
by textminer 13y ago
This was inspiring in a way many "how I started a company" posts aren't. I also appreciated the transparency re: the retention package offered to him in the Powerset acquisition. As a younger engineer, I've spent a lot of time focusing on s
71.
by textminer 13y ago
Why is so much of it non-fat?! The 2% milk fat Greek yogurt is so much richer, adds a trivial number of calories, and requires almost no sugar to taste "good".
72.
by textminer 13y ago
Isn't that less the externalities than the pool of folks using the existing hotel infrastructure (bending supply/demand) rather than exploring AirBnb/couch surfing/etc. I imagine if business travelers and tourists used AirBnB en masse, the
73.
by textminer 13y ago
Am I the only one driven crazy by reporting that states an ambiguous "average" for something asymmetric like salary, when the choice between median and mean as a measure of centrality makes all the difference in the world?
74.
by textminer 13y ago
Over a Fast Fourier Transform?! That's like saying an SVD or a decision tree constitutes a trade secret.
75.
by textminer 13y ago
The argument against testing seems to be "the Spike", when you're learning a language/concept/framework or are testing out ideas. But startups (at least the two I've worked for the last two years) seem way more heavily invested in the spike
76.
by textminer 13y ago
Only because it'd spectacularly fail.
77.
by textminer 13y ago
I don't think it'd be the folks living in a paradise. It'd be the folks fleeing persecution or economic horror. Why'd folks leave Europe for the New World, or any other migration? (And I mean the early ones-- not the gold rush-thinkin' ones
78.
by textminer 13y ago
Are you sure on the first part? My read of the historical narrative is that folks attribute the success of Apollo to JFK, not Nixon, and not Johnson. However wrong or right. (It should be pointed out that I was born in the mid-80s, and am f
79.
by textminer 13y ago
I fully agree. But only because an American colony on Mars would be aesthetically preferable to a SpaceX corporate colony on Mars, where colonists'd predictably work fifteen year sentences as indentured servant miners to pay off their voyag
80.
by textminer 13y ago
And even if you only see a few people richer than you, in a place with a high Gini coefficient, they're probably much richer than you. Human brains like to linearly extrapolate, so you probably assume that that non-linear explosion in growt
81.
by textminer 13y ago
Where's the FUD about recent college grads in the 1960's interviewing for jobs with full, unkempt beards, and making allusions to radical politics?
82.
by textminer 13y ago
(Just in case my comment sounded parodic because of my self-description, or contrarian because of your call for public bravery, it wasn't meant to be either.)
83.
by textminer 13y ago
I find your comments to be spectacularly useful. I'm a twenty something machine learning developer with a functional bent and high expectations for rewarding stimulation in VC-land, and you've recapitulated many of my own thoughts and concl
84.
by textminer 13y ago
If using raw bag of words or n-grams, why not hash the strings to maybe 2^13-1 slots, with something like MurmurHash3 or with multiple hashes to prevent collisions, then use that sparse vector as input to a deep learning model?
85.
by textminer 13y ago
I don't know. Beach pictures, drag queens (Shangela jumps out of a box!), cute kids, kitty cats, and club photos basically describes what I see the service as being used for by my chums. Not a bad diversion for keeping up with far-flung fri
86.
by textminer 13y ago
Don't mean to poo-poo. I love scrambling for models and methods. In the latest problem I worked on, I did the same thing. Implementing papers, trying to avail myself of semi-supervised or reduced-dimension representations, tweaking models a
87.
by textminer 13y ago
If I recall, BellKor made many, many models based on Gradient Boosted Decision Trees, Restricted Boltzmann Machines, and kNN. They tried many different feature subsets, added temporal weighting, and tried many reduced-dimensionality represe
88.
by textminer 13y ago
"Did you fall off a space elevator or something?!" Only because it might compel some contrarian to build a space elevator. And I really want a space elevator.
89.
by textminer 13y ago
Can't stress the importance of a REPL enough. The hands-on interactivity, call and response, and introspection pushes your understanding of the language (and ability to build out astounding new ideas) so much more quickly. Also love that he
90.
by textminer 14y ago
I love this entry. Two months ago, I wrote a perhaps more-pedantically-technical entry about the hash trick and its use in building a simple classifier, even making it play nice with Scipy's sparse linear algebra (eventually writing my own

More ›