sweezyjeezy

46 ms·

481.
by sweezyjeezy 11y ago
Seems a bit weak - I don't think you can really drill down onto one episode as one that 'hooks' a viewer, rather than all the subsequent episodes being excellent. At least, this article didn't convince me.
482.
Show HN: Animations containing every 24-bit RGB colour once(swarbrickjones.wordpress.com)
1 points by sweezyjeezy 11y ago|0 comments
483.
All 24-bit colours in one animation(swarbrickjones.wordpress.com)
2 points by sweezyjeezy 11y ago|0 comments
484.
by sweezyjeezy 11y ago
My pedantry is involuntary, these rules were drilled into me as a kid. When I see a mistake my eye gets snagged on it, as if I were running and stumbled over - it's annoying. So this article doesn't help me much personally.
485.
by sweezyjeezy 11y ago
If that exact story had happened to a person, it wouldn't be on the news in the West. I don't think people care more about animals more than humans, but I think it's very hard for humans to truly care all that much about hum
486.
by sweezyjeezy 11y ago
> These 9 people will be responsible for giving away $80 million per year in prizes. In the scientific world, that’s a huge program. It's about 0.25% of the NIH's investment budget. Doesn't sound like that big a deal to m
487.
by sweezyjeezy 11y ago
You can get something like 5% error rate on MNIST with a well tuned linear classifier, it's not comparable with ImageNet. Note that the computer vision techniques used before convnets used things like SIFT features, which are another w
488.
by sweezyjeezy 11y ago
I don't see how 'lazy' is a good word to use though. I would imagine someone who is lazy would be much less likely to write documentation than someone who isn't. Prudent is a much more fitting, if less eye-catching wor
489.
by sweezyjeezy 11y ago
MNIST is really easy, it's not a useful benchmark any more.
490.
by sweezyjeezy 11y ago
The image stuff works because we have found a way to model a good prior for it : convolution layers are basically enforcing some positional invariance and locality constraints on what our model believes the world looks like. Without this v
491.
by sweezyjeezy 11y ago
We've all but nailed the image recognition problem now though.
492.
by sweezyjeezy 11y ago
You're assuming increased height only leads to a stretch in the z axis. If you linearly scaled someone up from 5'3 to 5'4 in every direction, the increase in volume would be 5%. So maybe not the main factor, but definitely
493.
by sweezyjeezy 11y ago
This calculation is not exactly scientific, but : if U.S. women were an inch shorter then, that would make them 63 inches tall, instead of 64 inches (5'4). This equates to a 1.6% increase in height. If we assume that they have grown
494.
by sweezyjeezy 11y ago
Yeah the training set is called Imagenet, it's widely used in research.
495.
by sweezyjeezy 11y ago
Two things - firstly you need to understand the concept of overfitting : http://en.wikipedia.org/wiki/Overfitting . If teams were allowed to train on the full dataset, it would be possible to get a 100% score, yet sti
496.
by sweezyjeezy 11y ago
Not quite, it stacks neighbouring columns with the same pattern of colours on top of each other, e.g. http://i.imgur.com/35N4tPu.png The columns have to be the same pattern exactly (same height, same colours)
497.
by sweezyjeezy 11y ago
Since RNNs are self connected, in some sense it has a lot more depth than just 2, so they can learn very non-linear features (but the weights of the connections are shared across the layers).
498.
by sweezyjeezy 11y ago
The AI of the computer player is an RNN, which is a type of neural network. The objective of the game, what the controls do, and strategies were never hard coded into the AI, it learned them through trial and error, the way a human would.
499.
by sweezyjeezy 11y ago
You're missing the point entirely. You can explain why people might find it funny, that's the point.
500.
by sweezyjeezy 11y ago
I laughed when the author listed the obvious stopwords that always come high in these kinds of things, and then instead of just removing them, actually tried to imply that they are meaningful. I think it would be interesting to do some anal
501.
by sweezyjeezy 12y ago
That in no way explains why these K3 surfaces have anything to do with mock theta functions...
502.
by sweezyjeezy 12y ago
I heard about this dataset recently through the automatic statistician project - http://www.automaticstatistician.com/ Here is the analysis automatically produced by them : http://www.automaticstatistician.com&#x
503.
by sweezyjeezy 12y ago
But his first point still stands - none of these optimisations are very important.