Abstract | We survey learning algorithms for recurrent neural networks with hidden units, and put the various techniques into a common framework. We discuss xedpoint learning algorithms, namely recurrent backpropagation and deterministic Boltzmann Machines, and non- xedpoint algorithms, namely backpropagation through time, Elman's history cuto, and Jordan's output feedback architecture. Forward propagation, an online technique that uses adjoint equations, and variations thereof, are also discussed. In many cases, the uni ed presentation leads to generalizations of various sorts. We discuss advantages and disadvantages of temporally continuous neural networks in contrast to clocked ones, continue with some \tricks of the trade" for training, using, and simulating continuous time and recurrent neural networks. We present somesimulations, and at the end, address issues of computational complexity and learning speed.

Exploration plays a fundamental role in any active learning system. This study evaluates the role of exploration in active learning and describes several local techniques for exploration in finite, discrete domains, embedded in a reinforcement learning framework (delayed reinforcement). This paper distinguishes between two families of exploration schemes: undirected and directed exploration. While the former family is closely related to random walk exploration, directed exploration techniques memorize exploration-specific knowledge which is used for guiding the exploration search. In many finite deterministic domains, any learning technique based on undirected exploration is inefficient in terms of learning time, i.e. learning time is expected to scale exponentially with the size of the state space (Whitehead, 1991b) . We prove that for all these domains, reinforcement learning using a directed technique can always be performed in polynomial time, demonstrating the important role of e...

. To operate successfully in indoor environments, mobile robots must be able to localize themselves. Most current localization algorithms lack flexibility, autonomy, and often optimality, since they rely on a human to determine what aspects of the sensor data to use in localization (e.g., what landmarks to use). This paper describes a learning algorithm, called BaLL, that enables mobile robots to learn what features/landmarks are best suited for localization, and also to train artificial neural networks for extracting them from the sensor data. A rigorous Bayesian analysis of probabilistic localization is presented, which produces a rational argument for evaluating features, for selecting them optimally, and for training the networks that approximate the optimal solution. In a systematic experimental study, BaLL outperforms two other recent approaches to mobile robot localization. Keywords: artificial neural networks, Bayesian analysis, feature extraction, landmarks, localization, mobi...

We present a promising new mathematical method for tracking a user's pose (position and orientation) for interactive computer graphics. The method, which is applicable to a wide variety of both commercial and experimental systems, improves accuracy by properly assimilating sequential observations, filtering sensor measurements, and by concurrently autocalibrating source and sensor devices. It facilitates user motion prediction, multisensor data fusion, and higher report rates with lower latency than previous methods. Tracking systems determine the user's pose by measuring signals from low-level hardware sensors. For reasons of physics and economics, most systems make multiple sequential measurements which are then combined to produce a single tracker report. For example, commercial magnetic trackers using the SPASYN ( Space Synchro) system sequentially measure three magnetic vectors and then combine them mathematically to produce a report of the sensor pose. Our new approach produces tracker reports as each new lowlevel sensor measurement is made rather than waiting to form a complete collection of observations. Because single observations under-constrain the mathematical solution, we refer to our approach as single-constraint-at-a-time or SCAAT tracking. The key is that the single observations provide some information about the user's state, and thus can be used to incrementally improve a previous estimate. We recursively apply this principle, incorporating new sensor data as soon as it is measured. With this approach we are able to generate estimates more frequently, with less latency, and with improved accuracy. We present results from both an actual implementation, and from extensive simulations.

In this paper we describe a new approach for generalised nonlinear filtering. We show that the technique is more accurate, more stable, and far easier to implement than an extended Kalman filter. Several examples are provided, including the application of the new filter to problems involving discontinuous functions. 1 Introduction Possibly the most important problem arising in tracking and control applications is the representation and maintenance of uncertainty. The state of a system, whether measured or estimated, is rarely known perfectly because (a) measuring instruments and processes have limited precision, and/or (b) estimates of evolving systems are based on process models that fail to include all governing parameters. The uncertainty associated with a state estimate can be represented most generally by a probability distribution incorporating all knowledge about the state. Because the amount of knowledge about the state is inherently finite, a complete parameterisation of the ...

this paper, we describe a hierarchical network model of visual recognition that explains these experimental observations by using a form of the extended Kalman filter as given by the Minimum Description Length (MDL) principle. The model dynamically combines input-driven bottom-up signals with expectation-driven top-down signals to predict current recognition state. Synaptic weights in the model are adapted in a Hebbian manner according to a learning rule also derived from the MDL principle. The resulting prediction/learning scheme can be viewed as implementing a form of the Expectation-Maximization (EM) algorithm. The architecture of the model posits an active computational role for the reciprocal connections between adjoining visual cortical areas in determining neural response properties. In particular, the model demonstrates the possible role of feedback from higher cortical areas in mediating neurophysiological effects due to stimuli from beyond the classical receptive field. Si

Autonomous robots must be able to learn and maintain models of their environments. Research on mobile robot navigation has produced two major paradigms for mapping indoor environments: grid-based and topological. While grid-based methods produce accurate metric maps, their complexity often prohibits efficient planning and problem solving in large-scale indoor environments. Topological maps, on the other hand, can be used much more efficiently, yet accurate and consistent topological maps are often difficult to learn and maintain in large-scale environments, particularly if momentary sensor data is highly ambiguous. This paper describes an approach that integrates both paradigms: grid-based and topological. Grid-based maps are learned using artificial neural networks and naive Bayesian integration. Topological maps are generated on top of the grid-based maps, by partitioning the latter into coherent regions. By combining both paradigms, the approach presented here gains advantages from both worlds: accuracy/consistency and efficiency. The paper gives results for autonomous exploration, mapping and operation of a mobile robot in populated multi-room environments.

Over the last years, particle filters have been applied with great success to a variety of state estimation problems. In this paper we present a statistical approach to increasing the efficiency of particle filters by adapting the size of sample sets during the estimation process.

The long-term goal of our field is the creation and understanding of intelligence. Productive research in AI, both practical and theoretical, benefits from a notion of intelligence that is precise enough to allow the cumulative development of robust systems and general results. This paper outlines a gradual evolution in our formal conception of intelligence that brings it closer to our informal conception and simultaneously reduces the gap between theory and practice. 1 Artificial Intelligence AI is a field in which the ultimate goal has often been somewhat ill-defined and subject to dispute. Some researchers aim to emulate human cognition, others aim at the creation of

m-RI-m-95-33 submitted in partial fulfiumtnr of the reqimlmts for the degm of