Evolutionary Network Minimization: Adaptive Implicit Pruning of Successful Agents

Ganon, Zohar; Keinan, Alon; Ruppin, Eytan

doi:10.1007/978-3-540-39432-7_34

Cited by 5 publications

(2 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…P10 was obtained by a process in which, after the evolution of a successful agent, its synapses are pruned using an evolutionary network minimization algorithm [9] that deletes synapses and modifies the weights of the remaining ones so as to produce a similar agent with smaller neurocontroller. Like S10, P10 is equipped with a neurocontroller composed of binary McCulloch-Pitts neurons, but only 14 recurrent synapses out of the 100 original ones are left after applying the minimization algorithm.…”

Section: The Eaa Environmentmentioning

confidence: 99%

Axiomatic Scalable Neurocontroller Analysis via the Shapley Value

et al. 2006

Self Cite

View full text Add to dashboard Cite

One of the major challenges in the field of neurally-driven evolved autonomous agents is deciphering the neural mechanisms underlying their behavior. Aiming at this goal, we have developed the Multi-perturbation Shapley value Analysis (MSA) -the first axiomatic and rigorous method for deducing causal function localization from multiple perturbations data, substantially improving on earlier approaches. Based on fundamental concepts from game theory, the MSA provides an intuitive and formal way of definining and quantifying the contributions of network elements, as well as the functional interactions between them. The previously presented versions of the MSA require full knowledge (or at least an approximation) of the network's performance under all possible multiperturbations, making them applicable only for systems with a relatively small number of elements. This paper focuses on presenting new scalable MSA variants, allowing for the analysis of large complex networks in an efficient manner, including large-scale neurocontrollers. The successful operation of the MSA along with the new variants is demonstrated in the analysis of several neurocontrollers solving a food foraging task, consisting of up to 100 neural elements.

show abstract

Section: The Eaa Environmentmentioning

confidence: 99%

Axiomatic Scalable Neurocontroller Analysis via the Shapley Value

et al. 2006

Self Cite

View full text Add to dashboard Cite

show abstract

“…In the elimination step, each synaptic connection is eliminated (zeroed) with a certain fixed deletion probability. Other elimination criterions can also be used during the elimination step, such as the one we have previously introduced [5], according to which a synaptic connection is eliminated if the magnitude of its weight is smaller than a predefined threshold. Eliminated synapses remain zeroed and cannot be revived by mutation, thus creating pressure for network minimization.…”

Section: The Enm Algorithmmentioning

confidence: 99%

Neurocontroller Analysis via Evolutionary Network Minimization

Ganon¹,

Keinan

Ruppin

2006

Artificial Life

Self Cite

View full text Add to dashboard Cite

This study presents a new evolutionary network minimization (ENM) algorithm. Neurocontroller minimization is beneficial for finding small parsimonious networks that permit a better understanding of their workings.The ENM algorithm is specifically geared to an evolutionary agents setup, as it does not require any explicit supervised training error, and is very easily incorporated in current evolutionary algorithms. ENM is based on a standard genetic algorithm with an additional step during reproduction in which synaptic connections are irreversibly eliminated. It receives as input a successfully evolved neurocontroller and aims to output a pruned neurocontroller, while maintaining the original fitness level. The small neurocontrollers produced by ENM provide upper bounds on the neurocontroller size needed to solve a given task successfully, and can provide for more efficient hardware implementations.

show abstract

Evolutionary Supervised Machine Learning

Miikkulainen

2023

Genetic and Evolutionary Computation

View full text Add to dashboard Cite

Evolutionary Network Minimization: Adaptive Implicit Pruning of Successful Agents

Cited by 5 publications

References 6 publications

Axiomatic Scalable Neurocontroller Analysis via the Shapley Value

Axiomatic Scalable Neurocontroller Analysis via the Shapley Value

Neurocontroller Analysis via Evolutionary Network Minimization

Evolutionary Supervised Machine Learning

Contact Info

Product

Resources

About