gabrielgoh

10 ms·

31.
by gabrielgoh 9y ago
This is very fascinating, so thank you for looking into this. I think you may have misunderstood the rating system. You said >>>Let’s say I really have to produce a rating, though. What would I do? I would probably: - Find some exp
32.
by gabrielgoh 9y ago
where else would the training data come from?
33.
Conversations as two player games – teaching machines to negotiate with rollouts [pdf](s3.amazonaws.com)
3 points by gabrielgoh 9y ago|0 comments
34.
by gabrielgoh 9y ago
A lot of the problems with mathematical notation, I think, are solved problems in programming languages. Mathematicians could learn a lot from programmers in this regard - Variable Scope. In mathematical writing, the convention is that ever
35.
by gabrielgoh 9y ago
hey, I authored the first article :D
36.
by gabrielgoh 9y ago
I'm almost completing my phd in math, and as much as it pains me to say this, I think a great amount of deep learning can be accessible without "the math". And as tempting as it is to tell everyone to pay their mathematical d
37.
by gabrielgoh 9y ago
i fail to see whats mysterious about knowing something without direct proof. You could put a literal orange cube in a box, ship it across the galaxy, and you'd "know" anyone opening the box would be an orange cube. You don&#x
38.
by gabrielgoh 9y ago
Could you provide a baseline case of a complete random musical style? I know it will sound terrible, but I may be too much of a musical pleb to even notice the difference.
39.
by gabrielgoh 9y ago
I enjoyed the section on negative examples, and the things that don't work out. Makes me feel a lot better about the ton of things i've tried which didn't quite pan out the way I hoped.
40.
by gabrielgoh 9y ago
im not sure this is a good idea (as a consumer) anything which injects extra buying decisions in my life seems like a bad idea. Imagine having to wonder if I really wanted to spend that 0.01 cents on the next page of popular mechanics or no
41.
by gabrielgoh 9y ago
in a way, life is just a simple game with tricky combinatorics, and fairly trivial to attack with the tools of unsupervised learning.
42.
by gabrielgoh 9y ago
I love this visualization - but I think there's a very different intuition you get from PCA in high dimensions. I prefer to think of the singular vectors in PCA as an ordering of "prototype signals" for which some linear comb
43.
by gabrielgoh 9y ago
maybe this could be a qualitative way of understanding the body's insulin response to a meal, and to adjust one's diet accordingly (more meals, lower GI foods, etc)?
44.
by gabrielgoh 9y ago
it appears to be a way to measure complex behaviors, and higher cognitive functions, however - so I don't think the commenters are that far off.
45.
by gabrielgoh 10y ago
All I see here are your opinions, while the article has numerous citations and data to back up its claims. The reason why the article is interesting is because it goes against the common wisdom "you need to know the rules to break em&q
46.
by gabrielgoh 10y ago
Schmidhuber has forgotten to cite Leibniz. The results in this document are nothing more than a trivial corollary of the principle of metaphysical efficiency, not to mention pre-established harmony, which Leibniz has discovered the most gen
47.
by gabrielgoh 10y ago
to code a machine learning library from scratch, you must first invent the universe
48.
by gabrielgoh 10y ago
I think the distinction here between "handing over your data" and "letting a model train on the data on your device" may be more subtle than you might think. There is still no guarantee of privacy - it is trivial to cons
49.
by gabrielgoh 10y ago
They can - because the examples weren't labelled. In principle, you could use the model to train a whole lot of other things (what age group does a product appeal to? Is this product targeted at men/women?) with just a few hundred
50.
by gabrielgoh 10y ago
Its not quite the same. SOR is closer to coordinate descent in the way it acts
51.
by gabrielgoh 10y ago
These are valid criticisms, thank you very much for this.
52.
by gabrielgoh 10y ago
this is a good way to think about this.
53.
by gabrielgoh 10y ago
Very good question! I have considered this issue too. This form of weighting is the kind used in ADAM, and is qualitatively different from the updates described here. The tools of analysis in this article can be used to understand that iter
54.
by gabrielgoh 10y ago
Good question! The parameter has been set to a touch below the optimum. Your observation is accurate, there is indeed such a tradeoff, though it is less than you might think. The qualitative behavior of the system is very sensitive to chang
55.
by gabrielgoh 10y ago
I just mean p_i(\xi) = \xi^{i-1}. The notation is a little cumbersome here, but it pays of in the second equation
56.
by gabrielgoh 10y ago
Author here - I believe the problem of a "stiff system" you're referring to is exactly the problem of pathological curvature! Some points not touched on in the article. If the individual dimensions are of different scales, th
57.
by gabrielgoh 10y ago
the commenters below are correct - but I will push a change for this right away! It is my fault for not introducing it.
58.
by gabrielgoh 10y ago
Author here - I've created all the diagrams, though I've received really helpful editorial input from Shan Carter and Chris Olah. If you feel like doing some archeology, you can see for yourself the really ugly drafts in the githu
59.
by gabrielgoh 10y ago
This is indeed related. See the section on polynomial regression!
60.
by gabrielgoh 10y ago
Author here - yes! It can be used to solve symmetric PSD Systems, as the solution of Ax = b is also the minimizer of 0.5*x'Ax - b'x. Conjugate gradient can be seen as a special form of momentum with adaptive alphas and betas

More ›