Learning in compressed space

BOLeRo: Behavior optimization and learning for robots

Reinforcement learning and behavior optimization are becoming more and more popular in the field of robotics because algorithms are mature enough to tackle real problems in this domain. Robust implementations of state-of-the-art algorithms are often not publicly available though, and experiments are hardly reproducible because open-source implementations are often not available or are still in a stage of research code. Consequently, often it is infeasible to deploy these algorithms on robotic systems. BOLeRo closes this gap for policy search and evolutionary algorithms by delivering open-source implementations of behavior learning algorithms for robots. It is easy to integrate in robotic middlewares and it can be used to compare methods and develop prototypes in simulation.

Bootstrapping of Parameterized Skills Through Hybrid Optimization in Task and Policy Spaces

Modern robotic applications create high demands on adaptation of actions with respect to variance in a given task. Reinforcement learning is able to optimize for these changing conditions, but relearning from scratch is hardly feasible due to the high number of required rollouts. We propose a parameterized skill that generalizes to new actions for changing task parameters, which is encoded as a meta-learner that provides parameters for task-specific dynamic motion primitives. Our work shows that utilizing parameterized skills for initialization of the optimization process leads to a more effective incremental task learning. In addition, we introduce a hybrid optimization method that combines a fast coarse optimization on a manifold of policy parameters with a fine grained parameter search in the unrestricted space of actions. The proposed algorithm reduces the number of required rollouts for adaptation to new task conditions. Application in illustrative toy scenarios, for a 10-DOF planar arm, and a humanoid robot point reaching task validate the approach.

Modeling virtual organizations with Latent Dirichlet Allocation: a case for natural language processing

This paper explores a variety of methods for applying the Latent Dirichlet Allocation (LDA) automated topic modeling algorithm to the modeling of the structure and behavior of virtual organizations found within modern social media and social networking environments. As the field of Big Data reveals, an increase in the scale of social data available presents new challenges which are not tackled by merely scaling up hardware and software. Rather, they necessitate new methods and, indeed, new areas of expertise. Natural language processing provides one such method. This paper applies LDA to the study of scientific virtual organizations whose members employ social technologies. Because of the vast data footprint in these virtual platforms, we found that natural language processing was needed to 'unlock' and render visible latent, previously unseen conversational connections across large textual corpora (spanning profiles, discussion threads, forums, and other social media incarnations). We introduce variants of LDA and ultimately make the argument that natural language processing is a critical interdisciplinary methodology to make better sense of social 'Big Data' and we were able to successfully model nested discussion topics from forums and blog posts using LDA. Importantly, we found that LDA can move us beyond the state-of-the-art in conventional Social Network Analysis techniques.