tansey

11 ms·

61.
by tansey 13y ago
I really like this tool. It also helps highlight one thing that I think Sal really underplayed: it doesn't necessarily matter what long-term historical rates have been. If you assume that Silicon Valley is going to be hot for even just
62.
by tansey 13y ago
Well it's also unrealistic to buy a house for $750K in Mountain View. I think his point was that IF the house went for $750K, you could reasonably expect rent to be around $3K. That seems right to me-- a 1BR in SF sells and rents for a
63.
by tansey 13y ago
I'm not sure what PhD students you're hanging around, but all the ones I know are painfully aware of the fact that we're making terrible career choices. No one expects to get a job as a professor right out of the program, e
64.
by tansey 13y ago
Looks great! Do you have full support for numpy, scipy, matplotlib, pymc, and pandas?
65.
by tansey 13y ago
> Katy Jordan did a study of MOOCs recently [0], and it was found that the completion rates are in fact quite abysmal. I suspect that the reason for this is because there isn't much incentive to continue studying. No, it's be
66.
by tansey 13y ago
Exactly. Hyperloop in particular has the worst of all worlds. 1. Huge investment required up front. 2. First of its kind, so lots of risk. 3. Small potential gain relative to the investment. In fact, Hyperloop was never proposed as a VC fun
67.
by tansey 13y ago
You can't short an IPO...
68.
by tansey 13y ago
It's a great idea, but the site is really tough to use and when I first signed up, the app crashed. It seems really focused on NYC-- I couldn't add any coffee shops here in Austin. Also, how do I browse through people in my city?
69.
by tansey 13y ago
I tend to think of doctors as mechanics. Often when something breaks on my car, the explanation is "it's an old car, that happens eventually." And when they talk about it after fixing it, it's usually "Toyota's
70.
by tansey 13y ago
Looks like Tom hasn't actually blogged in 2 years though. That's not really very regular...
71.
by tansey 13y ago
> Users won't input that data. Sites like MyFitnessPal and MyNetDiary would likely disagree. They have hundreds of thousands of users. However, I agree that if you're in generally good health then it probably is too high of a b
72.
by tansey 13y ago
I think the oversimplified nutrition analytics is actually killing one of the biggest potential sources of insight from personal analytics. The key then is having a large, accurate database of nutritional information so you can see which fu
73.
by tansey 13y ago
Unfortunately, it's pretty tough to do it as just one person. All of my models assume you have a population of patients that are tracking their diets. With one person, you probably are best off just tracking your foods, medications, an
74.
by tansey 13y ago
Interesting. My PhD research is on precisely this topic-- nutrition diaries + machine learning for identifying the effect of foods on people with chronic illnesses, such as Crohn's or MS. I think this site is a good idea. It'd be
75.
by tansey 13y ago
Do NOT do things because they may make you rich. When people say that kind of thing, I always imagined doing jobs I would hate, but that wasn't the case. Building machine learning models to play poker, doing quant work for a hedge
76.
First legal online poker site goes live in Nevada(reuters.com)
2 points by tansey 13y ago|1 comments
77.
by tansey 14y ago
That was the same thought I had. The worst part is that they split it up into "chapters" which are really just one page of text and a pic each. The whole point of these designs is to be able to scroll down through the whole story like you w
78.
by tansey 14y ago
I've heard some pretty terrible stories about the vegetarian options offered by ZeroCater. It seems that many times they give the omnivores a full meal and then the vegetarian meal is the same meal without the meat, meaning it contains virt
79.
by tansey 14y ago
I saw a talk from Andrew Ng a couple of weeks ago where he was asked about the low retention rate. He countered that 40% of people who submit the first homework actually complete the class. I would say that is pretty great retention rate. P
80.
by tansey 14y ago
Indeed. If you are an undergrad looking to become a data scientist or even wanting to found a big data, machine learning, etc. company, starting a company with a BS in CS is not a better path than going to grad school.
81.
Back to the USSR by Way of Twitter(etilevich.wordpress.com)
2 points by tansey 14y ago|0 comments
82.
by tansey 14y ago
Isn't it just a bandit problem? Couldn't you use something like Thompson sampling to sort the comments? It'd have the nice benefits of adding random noise like reddit tries to do, while still having finite-time optimality guarantees.
83.
by tansey 14y ago
If WordGap was written in pure Python, without the use of any matrix libraries like numpy/scipy, then I think you could have done better. 7x speedup is actually pretty low for algorithms that I've rewritten to use numpy.
84.
How to Make Tough Decisions Objectively(wesleytansey.com)
2 points by tansey 14y ago|0 comments
85.
by tansey 14y ago
Right, I totally understand that. It is surprisingly difficult to find 2 hours a week, consistently every week, to sit down and work on one of these courses. I am just saying that the game theory course is by far the most accommodating of a
86.
by tansey 14y ago
I am a machine learning researcher working on game theory problems, so I suppose it's possible. I've never actually had any classes in game theory, so I took the Coursera one to have a firmer background on the subject. I actually learned a
87.
by tansey 14y ago
I agree with the OP's point in general, but disagree with you on the game theory course. This was the first course I've ever completed on Coursera and I think it's precisely because the time requirement is very low. They don't have any proj
88.
by tansey 14y ago
The promise of deep learning is that you will not have to do feature engineering. You give it raw features, like the pixels in the image, and then it discovers the latent features as you train it. So in a very real sense, yes we are limited
89.
by tansey 14y ago
1. Basically. It means that the network has a hard time pulling itself in any direction since the weights in the deeper layers are never really adjusted by very much. 2. It's been a while since I read the paper, but I believe that the justi
90.
by tansey 14y ago
Nice write up. I gave a presentation on DBNs for my Neural Networks class in Fall 2011. If you'd like references to the relevant papers and some more details on the algorithms and applications, here are the slides: https://docs.google.com/

More ›