This paper describes how agents can learn an internal model of the world structurally by focusing on the problem of behavior-based articulation. We develop an on-line learning scheme -- the so-called mixture of recurrent neural net (RNN) experts -- in which a set of RNN modules becomes self-organized as experts on multiple levels in order to account for the different categories of sensory-motor flow which the robot experiences. Autonomous switching of activated modules in the lower level actually represents the articulation of the sensory-motor flow. In the meanwhile, a set of RNNs in the higher level competes to learn the sequences of module switching in the lower level, by which articulation at a further more abstract level can be achieved. The proposed scheme was examined through simulation experiments involving the navigation learning problem. Our dynamical systems analysis clarified the mechanism of the articulation; the possible correspondence between the articulation...

The prefrontal cortex has long been thought to subserve both working memory (the holding of information online for processing) and executive functions (deciding how to manipulate working memory and perform processing). Although many computational models of working memory have been developed, the mechanistic basis of executive function remains elusive, often amounting to a homunculus. This article presents an attempt to deconstruct this homunculus through powerful learning mechanisms that allow a computational model of the prefrontal cortex to control both itself and other brain areas in a strategic, task-appropriate manner. These learning mechanisms are based on subcortical structures in the midbrain, basal ganglia, and amygdala, which together form an actor-critic architecture. The critic system learns which prefrontal representations are task relevant and trains the actor, which in turn provides a dynamic gating mechanism for controlling working memory updating. Computationally, the learning mechanism is designed to simultaneously solve the temporal and structural credit assignment problems. The model’s performance compares favorably with standard backpropagation-based temporal learning mechanisms on the challenging 1-2-AX working memory task and other benchmark working memory tasks.

The frontal cortex and basal ganglia interact via a relatively well-understood and elaborate system of interconnections. In the context of motor function, these interconnections can be understood as disinhibiting or "releasing the brakes" on frontal motor action plans --- the basal ganglia detect appropriate contexts for performing motor actions, and enable the frontal cortex to execute such actions at the appropriate time. We build on this idea in the domain of working memory through the use of computational neural network models of this circuit. In our model, the frontal cortex exhibits robust active maintenance, while the basal ganglia contribute a selective, dynamic gating function that enables frontal memory representations to be rapidly updated in a task-relevant manner. We apply the model to a novel version of the continuous performance task (CPT) that requires subroutine-like selective working memory updating, and compare and contrast our model with other existing models and th...

Long Short-Term Memory (LSTM, Hochreiter & Schmidhuber, 1997) can solve numerous tasks not solvable by previous learning algorithms for recurrent neural networks (RNNs). We identify a weakness of LSTM networks processing continual input streams that are not a priori segmented into subsequences with explicitly marked ends at which the network's internal state could be reset. Without resets, the state may grow indenitely and eventually cause the network to break down. Our remedy is a novel, adaptive \forget gate" that enables an LSTM cell to learn to reset itself at appropriate times, thus releasing internal resources. We review illustrative benchmark problems on which standard LSTM outperforms other RNN algorithms. All algorithms (including LSTM) fail to solve continual versions of these problems. LSTM with forget gates, however, easily solves them in an elegant way.

Recurrent networks (crossreference Chapter 12) can, in principle, use their feedback connections to store representations of recent input events in the form of activations. The most widely used algorithms for learning what to put in short-term memory, however, take too much time to be feasible or do not work well at all, especially when minimal time lags between inputs and corresponding teacher signals are long. Although theoretically fascinating, they do not provide clear practical advantages over, say, backprop in feedforward networks with limited time windows (see crossreference Chapters 11 and 12). With conventional "algorithms based on the computation of the complete gradient", such as "Back-Propagation Through Time" (BPTT, e.g., [22, 27, 26]) or "Real-Time Recurrent Learning" (RTRL, e.g., [21]) error signals "flowing backwards in time" tend to either (1) blow up or (2) vanish: the temporal evolution of the backpropagated error ex

A novel hierarchical neural network architecture for sensory-motor learning and behavior generation is proposed. Two levels of forward model neural networks are operated on different time scales while parametric interactions are allowed between the two network levels in the bottom-up and top-down directions. The models are examined through experiments of behavior learning and generation using a real robot arm equipped with a vision system. The results of the learning experiments showed that the behavioral patterns are learned by self-organizing the behavioral primitives in the lower level and combining the primitives sequentially in the higher level. The results contrast with prior work

This paper presents reinforcement learning with a Long ShortTerm Memory recurrent neural network: RL-LSTM. Model-free RL-LSTM using Advantage### learning and directed exploration can solve non-Markovian tasks with long-term dependencies between relevantevents. This is demonstrated in a T-maze task, as well as in a di#cult variation of the pole balancing task. 1

We introduce a new learning rule for fully recurrent neural networks which we call Backpropagation-Decorrelation rule (BPDC). It combines important principles: one-step backpropagation of errors and the usage of temporal memory in the network dynamics by means of decorrelation of activations. The BPDC rule is derived and theoretically justified from regarding learning as a constraint optimization problem and applies uniformly in discrete and continuous time. It is very easy to implement, and has a minimal complexity of 2N multiplications per time-step in the single output case. Nevertheless we obtain fast tracking and excellent performance in some benchmark problems including the Mackey-Glass time-series.

Abstract not found

The temporal distance between events conveys information essential for numerous sequential tasks such as motor control and rhythm detection. While Hidden Markov Models tend to ignore this information, recurrent neural networks (RNNs) can in principle learn to make use of it. We focus on Long Short-Term Memory (LSTM) because it has been shown to outperform other RNNs on tasks involving long time lags. We find that LSTM augmented by "peephole connections" from its internal cells to its multiplicative gates can learn the fine distinction between sequences of spikes spaced either 50 or 49 time steps apart without the help of any short training exemplars. Without external resets or teacher forcing, our LSTM variant also learns to generate stable streams of precisely timed spikes and other highly nonlinear periodic patterns. This makes LSTM a promising approach for tasks that require the accurate measurement or generation of time intervals.