Solving Rubik’s Cube With a Robot Hand
OpenAI has trained a pair of neural networks to solve the Rubik’s Cube with a human-like robot hand. The neural networks are trained entirely in simulation, using the same reinforcement learning code as OpenAI Five paired with a new technique called Automatic Domain Randomization (ADR).
(OpenAI) / (Synced) / (Read the Paper) / (Watch all videos)
Giving Robots a Faster Grasp
MIT engineers have found a way to significantly speed up the planning process required for a robot to adjust its grasp on an object by pushing that object against a stationary surface. Whereas traditional algorithms would require tens of minutes for planning out a sequence of motions, the new team’s approach shaves this preplanning process down to less than a second.
Assembler Robots Make Large Structures From Little Pieces
“This paper is a treat,” says Aaron Becker, an associate professor of electrical and computer engineering at the University of Houston, who was not associated with this work. “It combines top-notch mechanical design with jaw-dropping demonstrations, new robotic hardware, and a simulation suite with over 100,000 elements,” he says.
Connections Between Support Vector Machines, Wasserstein Distance and Gradient-Penalty GANs
Researchers generalize the concept of maximum margin classifiers (MMCs) to arbitrary norms and non-linear functions. Support Vector Machines (SVMs) are a special case of MMC. They find that MMCs can be formulated as Integral Probability Metrics (IPMs) or classifiers with some form of gradient norm penalty.
On Empirical Comparisons of Optimizers for Deep Learning
In this paper, they demonstrate the sensitivity of optimizer comparisons to the metaparameter tuning protocol. The findings suggest that the metaparameter search space may be the single most important factor explaining the rankings obtained by recent empirical comparisons in the literature.
(Google Brain & DeepMind & Institute for Advanced Study)
Functional RL with Keras and Tensorflow Eager
Researchers explore a functional paradigm for implementing reinforcement learning (RL) algorithms. The paradigm will be that developers write the numerics of their algorithm as independent, pure functions, and then use a library to compile them into policies that can be trained at scale.
(Berkeley Artificial Intelligence Research)
You May Also Like
Using Conditional GANs to Build Zelda Game Levels
A team of researchers has proposed a Generative Adversarial Network based model tasked with creating playable and aesthetically appealing game levels for popular action-adventure video game series The Legend of Zelda.
Shake Your Booty: AI Deepfakes Dance Moves From a Single Picture
If you’ve ever secretly wished you could move your body like Joaquín Cortés — well, at least in a video — a new AI-powered 3D body mesh recovery module called Liquid Warping GAN can give you a leg up.
Global AI Events
October 23-25: NVIDIA’s GPU Technology Conference (GTC) in Munich, Germany
October 27-November 3: International Conference on Computer Vision (ICCV) in Seoul,South Korea
November 8: MLconf in San Francisco, United States
November 13-14: AI & Big Data Expo in Santa Clara, United States