From the Publisher:
The old opposition of matter versus mind stubbornly persists in the way we study mind and brain. In treating cognition as problem solving, Andy Clark suggests, we may often abstract too far from the very body and world in which our brains evolved to guide us. Whereas the mental has been treated as a realm that is distinct from the body and the world, Clark forcefully attests that a key to understanding brains is to see them as controllers of embodied activity. From this paradigm shift he advances the construction of a cognitive science of the embodied mind.

/pdf/being-there-putting-brain-body-and-world-together-again-31tt4y51ml.pdf

Being There: Putting Brain, Body, and World Together Again

An Introduction to MultiAgent Systems.

Learning and evolution are two fundamental forms of adaptation. There has been a great interest in combining learning and evolution with artificial neural networks (ANNs) in recent years. This paper: 1) reviews different combinations between ANNs and evolutionary algorithms (EAs), including using EAs to evolve ANN connection weights, architectures, learning rules, and input features; 2) discusses different search operators which have been used in various EAs; and 3) points out possible future research directions. It is shown, through a considerably large literature review, that combinations between ANNs and EAs can lead to significantly better intelligent systems than relying on ANNs or EAs alone.

/pdf/evolving-artificial-neural-networks-4kzrbu7pkn.pdf

Evolving artificial neural networks

https://users.dimi.uniud.it/~antonio.dangelo/Robotica/2012/helper/leggedRobot/Ijspeert08NN.pdf

2008 Special Issue: Central pattern generators for locomotion control in animals and robots: A review

Recurrent neural networks (RNNs) have been widely adopted in research areas concerned with sequential data, such as text, audio, and video. However, RNNs consisting of sigma cells or tanh cells are...

A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures

We would like the behavior of the artificial agents that we construct to be as well-adapted to their environments as natural animals are to theirs. Unfortunately, designing controllers with these properties is a very difficult task. In this article, we demonstrate that continuous-time recurrent neural networks are a viable mechanism for adaptive agent control and that the genetic algorithm can be used to evolve effective neural controllers. A significant advantage of this approach is that one need specify only a measure of an agent's overall performance rather than the precise motor output trajectories by which it is achieved. By manipulating the performance evaluation, one can place selective pressure on the development of controllers with desired properties. Several novel controllers have been evolved, including a chemotaxis controller that switches between different strategies depending on environmental conditions, and a locomotion controller that takes advantage of sensory feedback if available but th...

Evolving dynamical neural networks for adaptive behavior

For many evolvable hardware applications, small size and power efficiency are critical design considerations. One manner in which significant memory, and thus, power and space savings can be realized in a hardware-based evolutionary algorithm is to represent populations of candidate solutions as probability vectors rather than as sets of bit strings. The compact genetic algorithm (CGA) is a probability vector-based evolutionary algorithm that can be efficiently and elegantly implemented in digital hardware. Unfortunately, the CGA is a very weak, first order, evolutionary algorithm that is unlikely to possess sufficient search power to enable intrinsic evolvable hardware applications. In this paper, we further develop a number of modifications to the basic CGA that significantly improve its search efficacy without substantially increasing the size and complexity of its hardware implementation. The paper provides both benchmark results demonstrating increased efficacy and a conceptual data path/microcontroller design suitable for implementation in digital hardware. Following, it demonstrates efficient implementation by making a head-to-head comparison of field programmable gate array implementations of both the classic CGA and a member of our family of modifications. The paper concludes with a discussion of future research, including several additional extensions that we expect will further increase search efficacy without increasing implementation cost.

A family of compact genetic algorithms for intrinsic evolvable hardware

Application of evolved locomotion controllers to a hexapod robot

Are there general principles for pattern generation? We examined this question by analyzing the operation of large populations of evolved model central pattern generators (CPGs) for walking. Three populations of model CPGs were evolved, containing three, four, or five neurons. We identified six general principles. First, locomotion performance increased with the number of interneurons. Second, the top 10 three-, four-, and five-neuron CPGs could be decomposed into dynamical modules, an abstract description developed in a companion article. Third, these dynamical modules were multistable: they could be switched between multiple stable output configurations. Fourth, the rhythmic pattern generated by a CPG could be understood as a closed chain of successive destabilizations of one dynamical module by another. A combinatorial analysis enumerated the possible dynamical modular structures. Fifth, one-dimensional modules were frequently observed and, in some cases, could be assigned specific functional roles. Finally, dynamic dynamical modules, in which the modular structure itself changed over one cycle, were frequently observed. The existence of these general principles despite significant variability in both patterns of connectivity and neural parameters was explained by degeneracy in the maps from neural parameters to neural dynamics to behavior to fitness. An analysis of the biomechanical properties of the model body was essential for relating neural activity to behavior. Our studies of evolved model circuits suggest that, in the absence of other constraints, there is no compelling reason to expect neural circuits to be functionally decomposable as the number of interneurons increase. Analyzing idealized model pattern generators may be an effective methodology for gaining insights into the operation of biological pattern generators.

Evolution and analysis of model CPGs for walking: II. General principles and individual variability.

Can one develop an abstract description of the dynamics of pattern generators that provides quantitative insight into their operation? We explored this question by examining the dynamics of a model central pattern generator that was created using an evolutionary algorithm. We propose an abstract description based on the concept of a dynamical module, a set of neurons that simultaneously make their transitions from one quasistable state to another while the synaptic inputs that they receive from other neurons remain essentially constant, thus temporarily reducing the dimensionality of the circuit dynamics. Using the mathematical tools of dynamical systems theory, we describe a method for identifying dynamical modules and demonstrate that this concept can be used to quantitatively characterize constraints on neural architecture, account for phase durations, and predict the effects of parameter changes. Moreover, this abstract description reveals coordinated parameter changes that leave the overall circuit dynamics essentially unchanged. In a companion article we employ this abstract description to examine the relationship between general principles and individual variability in large populations of evolved model pattern generators.

John C. Gallagher

Papers

Evolving dynamical neural networks for adaptive behavior

A family of compact genetic algorithms for intrinsic evolvable hardware

Application of evolved locomotion controllers to a hexapod robot

Evolution and analysis of model CPGs for walking: II. General principles and individual variability.

Evolution and analysis of model CPGs for walking: I. Dynamical modules.