jostmey

60 ms·

661.
by jostmey 10y ago
The method I presented is for classification. It is not a generative model, although there's no reason why it can't be incorporated into a generative model. The RWA model could be used to perform sequence to sequence mappings like
662.
by jostmey 10y ago
I will try to update the README. The results are described in the arXiv link. I hope the paper is clear. Here's the link: https://arxiv.org/pdf/1703.01253.pdf
663.
by jostmey 10y ago
Author here. I'd be happy to address any questions. Constructive criticism is always welcome.
664.
New kind of recurrent neural network using attention(github.com)
155 points by jostmey 10y ago|42 comments
665.
by jostmey 10y ago
The US pays an unfair burden to cover the cost of developing new drugs. That said, many drug companies seem more concerned with selling drugs that are barely better than a placebo and have cut back on research efforts.
666.
by jostmey 10y ago
Author here: A typo exist in equation (2). The index in the denominator should be "j". Thanks to ChuckSeven for pointing this out. The document was originally written in word and then transcribed into latex
667.
New kind of RNN connects every step with recurrent weighted average(arxiv.org)
3 points by jostmey 10y ago|1 comments
668.
by jostmey 10y ago
I interpret this as Silicon valley losing its way. I'm not judging LSD use, but if you need it to get through the day then you need to find a different career. If you need your high paying, high stress job to afford your lifestyle, you
669.
by jostmey 10y ago
Maybe stated in a different way, how good do your priors have to be to learn and understand a language? With an infinite amount of training data, any language can be learned. The question is what prior assumptions are required to learn a la
670.
by jostmey 10y ago
I think it is important to match coworkers by their caliber. You're right, if you have one really amazing developer and a bunch of mediocre developers that one really good developer could become bitter.
671.
by jostmey 10y ago
Could CO be produced by a clogged clothing dryer?
672.
by jostmey 10y ago
I was just there this past week. I could not believe how much the city has changed since I was last there 5 years ago. It is taking off
673.
by jostmey 10y ago
Quote: "Driving a new (untrained) section of the Royal Raceway:" So the author did a proper test of the model by scoring it on an unseen track to make sure it generalizes! This is very awesome!
674.
by jostmey 10y ago
Quote: "many top hedge funds and trading shops came to NIPS to run career booths, but there was a surprising lack of interest from attendees compared to the likes of Apple, Facebook, Deepmind, Google, etc." Maybe Apple, Facebook,
675.
by jostmey 10y ago
I didn't believe it but there it is! What a typo...
676.
by jostmey 10y ago
Maybe we should be asking why some industries are heavily taxed and others are not
677.
by jostmey 10y ago
So facebook, which needs us to spend as much time on the website as possible to maintain advertising revenue, has the perverse incentive of keeping everyone single. So this could be the answer to the Fermi Paradox
678.
by jostmey 10y ago
Yeah, putting together the software to generate an image is not lot building a database or webservice. It requires engineering and science knowledge well outside the domain of computer science.
679.
by jostmey 10y ago
A bunch of creepos spying on people. Oh wait, its the government, so its not weird :-)
680.
by jostmey 10y ago
Quote "Sudden mobile phone battery draining" Yikes! It sounds like the camera and microphone had been activated remotely and were being use to spy on citizens .
681.
by jostmey 10y ago
The ending is a bit of a twist. I thought the student would encounter a broken man--most of the homeless I see are broken individuals--but not so.
682.
by jostmey 10y ago
Neurogensis? How about neural death as a way to prune large neural networks into more compact ones--now that is a research idea!
683.
by jostmey 10y ago
Apple just wants to make sure the public understands how little it respects its engineers.
684.
by jostmey 10y ago
Disclaimer, I haven't looked at their work to closely. I think the process they use to select the hyperparameters overfits on the labels in the validation set. The true size of the labelled data includes the 100 from the training set a
685.
by jostmey 10y ago
I think there are a lot of programmers try playing around with "deep learning" and it doesn't work for them. But they lack the knowledge necessary to make it work, such as calculus, statistics, signal processing theory, ect.
686.
by jostmey 10y ago
And such news services used to exist, and nobody paid for them once it became free on the Internet. I saw it happen as I grew up--my dad worked in a newsroom as a journalist.
687.
by jostmey 10y ago
too of much of my day is spent on stupid stuff. Time must be guarded carefully--cut out all the crap
688.
by jostmey 10y ago
You are spot on correct. And regenerating lost body parts will require far more than just flipping the activation of a few genes.
689.
by jostmey 10y ago
Why not KL-Divergence, which measures the error between a target distribution and the current distribution? From the perspective of Information Theory, it is the best error measurement. Oh, and let's not forget that for a lot of proble
690.
by jostmey 10y ago
On a side note I read that book too & loved it.

More ›