When a new-born animal (agent) opens its eyes, what it sees is a patchwork of light and dark patterns, the natural scene.What is perceived by the agent at this moment is based on the patternof neural spikes in its brain. Life-long learning begins with such a flood of spikes in the brain. All knowledge and skills learned by the agent are mediated by such spikes, thus it is critical to understand what information these spikes convey and how they can be used to generate meaningful behavior. Here, we consider how agents can autonomously understand the meaning of these spikes without direct reference to the stimulus. We find that this problem, the problem of grounding, is unsolvable if the agent is passively perceiving, and that it can be solved only through self-initiated action. Furthermore, we show that a simple criterion, combined with standard reinforcement learning, can help solve this problem. We will present simulation results and discuss the implications of these results on life-long learning.

Autonomous robots demand complex behavior to deal with unstructured environments. To meet these expectations, a robot needs to address a suite of problems associated with long term knowledge acquisition, representation, and execution in the presence of partial information. In this paper, we address these issues by the acquisition of broad, domain general skills using an intrinsically motivated reward function. We show how these skills can be represented compactly and used hierarchically to obtain complex manipulation skills. We further present a Bayesian model using the learned skills to model objects in the world, in terms of the actions they afford. We argue that our knowledge representation allows a robot to both predict the dynamics of objects in the world as well as recognize them.

In 2009, an average American spent 3 hours per day watching TV. Recent advances in TV entertainment technologies, such as on-demand content, browsing the Internet, and 3D displays, have changed the traditional role of the TV and turned it into the center of home entertainment. Most of these technologies are personal and would benefit from seamless identification of who sits in front of the TV. In this work, we propose a practical and highly accurate solution to this problem. This solution uses a camera, which is mounted on a TV, to recognize faces of people in front of the TV. To make the approach practical, we employ online learning on graphs and show that we can learn highly accurate face models in difficult circumstances from as little as one labeled example. To evaluate our solutions, we collected a 10-hour long dataset of 8 people who watch TV. Our precision and recall are in the upper nineties, and show the promise of utilizing our approach in an embedded setting.

Recently, information gain has been proposed as a candidate intrinsic motivation for lifelong learning agents that may not always have a specific task.  In the InfoMax control framework, reinforcement learning is used to find a control policy for a POMDP in which movement and sensing actions are selected to reduce Shannon entropy as quickly as possible. In this study, we implement InfoMax control on a robot which can move between objects and perform sound-producing manipulations on them.  We formulate a novel latent variable mixture model for acoustic similarities and learn InfoMax polices that allow the robot to rapidly reduce uncertainty about the categories of the objects in a room. We find that InfoMax with our improved acoustic model leads to policies which lead to high classification accuracy.  Interestingly, we also find that with an insufficient model, the InfoMax policy eventually learns to "bury its head in the sand" to avoid getting additional evidence that might increase uncertainty.  We discuss the implications of this finding for InfoMax as a principle of intrinsic motivation in lifelong learning agents.

Versatile agents situated in rich, dynamic environments must It is not necessarily possible to select the source knowledge be capable of continually learning and refining their knowledge to transfer to a new target task by examining only the surface through experience. These agents will face a variety of similarities between the tasks. The selection must support learning tasks, and can transfer knowledge between tasks to the process of knowledge transfer by choosing source improve performance and accelerate learning. In this context, knowledge based on whether it will transfer well to the target a learning task can be as simple as discovering the effects task. In our previous work, we developed methods that of an operator on the environment, or as complex as accomplishing identify the source knowledge to transfer based on this concept a specific goal -- anything that can be learned of transferability to the target task. Intuitively, transferability can be considered a task. As the agent experiences and learns is the amount that the transferred information is a model for each task, it gains access to new data and knowledge.

The paper addresses the following issues:  (a) how to represent semantic information from natural language so that a vision model can utilize it?  (b) how to extract the salient textual information relevant to vision?  For a given domain, we present a new model of semantic extraction that takes into account word relatedness as well as word disambiguation in order to apply to a vision model. We automatically process the text transcripts and perform syntactic analysis to extract dependency relations. We then perform semantic extraction on the output to filter semantic entities related to actions. The resulting data are used to populate a matrix of co-occurrences utilized by the vision processing modules.  Results show that explicitly modeling the co-occurrence of actions and tools significantly improved performance.

We study the theoretical advantages of active learning over passive learning. Specifically, we prove that, in noise-free classifier learning for VC classes, any passive learning algorithm can be transformed into an active learning algorithm with asymptotically strictly superior label complexity for all nontrivial target functions and distributions. We further provide a general characterization of the magnitudes of these improvements in terms of a novel generalization of the disagreement coefficient. We also extend these results to active learning in the presence of label noise, and find that even under broad classes of noise distributions, we can typically guarantee strict improvements over the known results for passive learning.

We propose an intelligent tutoring system that constructs a curriculum of hints and problems in order to teach a student skills with a rich dependency structure. We provide a template for building a multi-layered Dynamic Bayes Net to model this problem and describe how to learn the parameters of the model from data. Planning with the DBN then produces a teaching policy for the given domain. We test this end-to-end curriculum design system in two human-subject studies in the areas of finite field arithmetic and artificial language and show this method performs on par with hand-tuned expert policies.

Robots with many sensors are capable of generating volumes of high-dimensional perceptual data. Making sense of this data and extracting useful knowledge from it is a difficult problem. For robots lacking proper models, trying to understand a stream of uninterpreted data is an especially acute problem. One critical step in linking raw uninterpreted perceptual data to cognition is dimensionality reduction. Current methods for reducing the dimension of data do not meet the demands of a robot situated in the world, and methods that use only perceptual data do not take full advantage of the interactive experience of an embodied robot agent. This work proposes a new scalable, incremental and active approach to dimensionality reduction suitable for extracting geometric knowledge from uninterpreted sensors and effectors. The proposed method uses distinctive state abstractions to organize early sensorimotor experience and sensorimotor embedding to incrementally learn accurate geometric representations based on experience. This approach is applied to the problem of learning the geometry of sensors, space, and objects. The result is evaluated using techniques from statistical shape analysis.

Intelligent tutoring systems (ITSs) provide students with a one-on-one tutor, allowing them to work at their own pace, and helping them to focus on their weaker areas. The RUR1–Python Learning Environment (RUR-PLE), a game-like virtual environment to help students learn to program, provides an interface for students to write their own Python code and visualize the code execution (Roberge 2005). RUR-PLE provides a fixed sequence of learning lessons for students to explore. We are extending RUR-PLE to develop the Playing to Program (PtP) ITS, which consists of three components: (1) a Bayesian student model that tracks student competence, (2) a diagnosis module that provides tailored feedback to students, and (3) a problem selection module that guides the student’s learning process. In this paper, we summarize RUR-PLE and the PtP design, and describe an ongoing user study to evaluate the predictive accuracy of our student modeling approach.