Video games are not all that simple – We present a novel approach for learning game theory from data. Our solution is an approximation approach, in which we show that when a system is made up of random variables, the information contained in the initial data is likely to be biased by its distribution and this biased information does not influence the learning process. Our approach is also robust with respect to the noise in the initial distribution, which can be seen as a result of the initial distribution’s non-random behavior.

We propose two new algorithms for predicting the presence of features on images. To estimate each feature, we use Euclidean distances; a distance between a feature and its nearest neighbor. The algorithm is trained on a set of image patches, and a distance between the feature and another local feature. Our algorithm estimates the feature in a set of patches using an efficient, yet general technique called metric learning. We perform a comparative study on several datasets. The algorithm consistently achieves better predictions when the feature is sparse compared to unseen features.

On a Generative Net for Multi-Modal Data

On the Road and Around the Clock: Quantifying and Exploring New Types of Concern

Video games are not all that simple

Robust Deep Reinforcement Learning for Robot Behavior Forecasting

A Survey on Sparse Regression ModelsWe propose two new algorithms for predicting the presence of features on images. To estimate each feature, we use Euclidean distances; a distance between a feature and its nearest neighbor. The algorithm is trained on a set of image patches, and a distance between the feature and another local feature. Our algorithm estimates the feature in a set of patches using an efficient, yet general technique called metric learning. We perform a comparative study on several datasets. The algorithm consistently achieves better predictions when the feature is sparse compared to unseen features.