hardmaru

55 ms·

661.
Meta-Learning in 50 Lines of JAX(blog.evjang.com)
35 points by hardmaru 8y ago|0 comments
662.
This Waifu Does Not Exist(thiswaifudoesnotexist.net)
175 points by hardmaru 8y ago|23 comments
663.
Learning to Control Self-Assembling Morphologies: Generalization via Modularity(pathak22.github.io)
6 points by hardmaru 8y ago|0 comments
664.
Learning Latent Dynamics for Planning from Pixels(planetrl.github.io)
5 points by hardmaru 8y ago|0 comments
665.
Trump will launch American AI Initiative with an executive order(technologyreview.com)
4 points by hardmaru 8y ago|0 comments
666.
The Limitations of Deep Learning for Vision and How We Might Fix Them(thegradient.pub)
8 points by hardmaru 8y ago|0 comments
667.
Artificial Intelligence and the Art of Mario Klingemann(sothebys.com)
49 points by hardmaru 8y ago|9 comments
668.
Raspberry Pi Opens First High Street Store in Cambridge(bbc.com)
246 points by hardmaru 8y ago|149 comments
669.
POET: Endlessly Generating Increasingly Complex Environments and Their Solutions(eng.uber.com)
8 points by hardmaru 8y ago|0 comments
670.
The family of a boy without a cerebellum found out how to take its place(economist.com)
1 points by hardmaru 8y ago|0 comments
671.
A GAMEBOY supercomputer(towardsdatascience.com)
3 points by hardmaru 8y ago|0 comments
672.
Show HN: Kuzushiji-MNIST(github.com)
6 points by hardmaru 8y ago|1 comments
673.
Asking the right question is more important than getting the right answer(lemire.me)
87 points by hardmaru 8y ago|14 comments
674.
by hardmaru 8y ago
link to their paper: https://arxiv.org/abs/1812.03473
675.
Comixify: Transform video into a comics(comixify.ii.pw.edu.pl)
77 points by hardmaru 8y ago|8 comments
676.
Deep Learning for Classical Japanese Literature(arxiv.org)
1 points by hardmaru 8y ago|0 comments
677.
So you want to be a Research Scientist(medium.com)
2 points by hardmaru 8y ago|0 comments
678.
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks(gandissect.csail.mit.edu)
180 points by hardmaru 8y ago|18 comments
679.
China’s brightest children are being recruited to develop AI ‘killer bots’(scmp.com)
2 points by hardmaru 8y ago|0 comments
680.
Why a career in computer programming sucks (2007)(web.archive.org)
1 points by hardmaru 8y ago|0 comments
681.
Learning to Dress: Synthesizing Human Dressing Motion via Reinforcement Learning(cc.gatech.edu)
92 points by hardmaru 8y ago|1 comments
682.
by hardmaru 8y ago
There was also a nice discussion about Ghost Kanji a few months back: https://news.ycombinator.com/item?id=17637375 A spectre is haunting unicode: https://www.dampfkraft.com/ghost-characters.html
683.
Ten simple rules for scientists: Improving your writing productivity(journals.plos.org)
1 points by hardmaru 8y ago|0 comments
684.
by hardmaru 8y ago
Some more information about this project: blog: https://magenta.tensorflow.org/pianogenie pdf: https://arxiv.org/abs/1810.05246 youtube video: https://youtu.be/YRb0XAnUpIk
685.
Piano Genie(tensorflow.github.io)
5 points by hardmaru 8y ago|1 comments
686.
by hardmaru 8y ago
I’ll be releasing the code to reproduce the experiments soon.
687.
by hardmaru 8y ago
There’s only 8 params for the biped walker, compared to thousands of params for the policy network. In this case it should be possible to scale to the thousands of environment parameters. For instance if we want to learn the material config
688.
by hardmaru 8y ago
Hi, TL;DR Usually in reinforcement learning, the body designs of robot agents are fixed. Here, the agent can learn a task, and also learn a better design of its body at the same time. Link to the blog article: https://designrl.gi
689.
Reinforcement Learning for Improving Agent Design(designrl.github.io)
103 points by hardmaru 8y ago|6 comments
690.
Large Scale GAN Training for High Fidelity Natural Image Synthesis(arxiv.org)
2 points by hardmaru 8y ago|0 comments

More ›