Multiagent Learning with Spatiotemporal Context: Application to Predicting Consumer’s Behaviors – Deep generative models and object models are becoming increasingly popular for representing, modelling and learning new data. In this paper, we propose a novel approach for learning generic objects that do not involve the appearance of human faces. This approach consists in a hierarchical generative model. The model is trained using an unsupervised learning method, which relies on pre-trained models for the tasks in hand and then combines the results of three pre-trained generative models. The learned objects represent both the human faces and the faces of others. We show that the approach can learn object models by using a novel spatial-temporal connection that is based on latent-source representations such as the appearance of human faces in real world images. The experiments on real-world datasets demonstrate that the technique can significantly outperform unsupervised supervised classification methods on both real-world and synthetic datasets.

The use of natural language to help people understand, reason about and understand is a major issue in social science research. In this paper, we investigate whether or not natural language is a powerful tool for cognitive science assessment. We perform a series of experiments to evaluate the effectiveness and computational cost of natural language processing technologies, i.e. cognitive systems and cognitive processing systems. We present several results that show that natural language processing technologies can offer very substantial and efficient machine learning capabilities.

Evaluating the Accuracy of Text Trackers using the Inductive Logic Problem

Inference on Regression Variables with Bayesian Nonparametric Models in Log-linear Time Series

Multiagent Learning with Spatiotemporal Context: Application to Predicting Consumer’s Behaviors

Lipschitz Optimization for Feature Interpolation by Low-Rank Fusion of Gaussian and Joint Features

Fluency-based machine learning methods for the evaluation of legal textsThe use of natural language to help people understand, reason about and understand is a major issue in social science research. In this paper, we investigate whether or not natural language is a powerful tool for cognitive science assessment. We perform a series of experiments to evaluate the effectiveness and computational cost of natural language processing technologies, i.e. cognitive systems and cognitive processing systems. We present several results that show that natural language processing technologies can offer very substantial and efficient machine learning capabilities.