340 research outputs found
An Efficient Method for online Detection of Polychronous Patterns in Spiking Neural Network
Get PDFPolychronous neural groups are effective structures for the recognition of
precise spike-timing patterns but the detection method is an inefficient
multi-stage brute force process that works off-line on pre-recorded simulation
data. This work presents a new model of polychronous patterns that can capture
precise sequences of spikes directly in the neural simulation. In this scheme,
each neuron is assigned a randomized code that is used to tag the post-synaptic
neurons whenever a spike is transmitted. This creates a polychronous code that
preserves the order of pre-synaptic activity and can be registered in a hash
table when the post-synaptic neuron spikes. A polychronous code is a
sub-component of a polychronous group that will occur, along with others, when
the group is active. We demonstrate the representational and pattern
recognition ability of polychronous codes on a direction selective visual task
involving moving bars that is typical of a computation performed by simple
cells in the cortex. The computational efficiency of the proposed algorithm far
exceeds existing polychronous group detection methods and is well suited for
online detection.Comment: 17 pages, 8 figure
Bi-fidelity Evolutionary Multiobjective Search for Adversarially Robust Deep Neural Architectures
Full text linkDeep neural networks have been found vulnerable to adversarial attacks, thus
raising potentially concerns in security-sensitive contexts. To address this
problem, recent research has investigated the adversarial robustness of deep
neural networks from the architectural point of view. However, searching for
architectures of deep neural networks is computationally expensive,
particularly when coupled with adversarial training process. To meet the above
challenge, this paper proposes a bi-fidelity multiobjective neural architecture
search approach. First, we formulate the NAS problem for enhancing adversarial
robustness of deep neural networks into a multiobjective optimization problem.
Specifically, in addition to a low-fidelity performance predictor as the first
objective, we leverage an auxiliary-objective -- the value of which is the
output of a surrogate model trained with high-fidelity evaluations. Secondly,
we reduce the computational cost by combining three performance estimation
methods, i.e., parameter sharing, low-fidelity evaluation, and surrogate-based
predictor. The effectiveness of the proposed approach is confirmed by extensive
experiments conducted on CIFAR-10, CIFAR-100 and SVHN datasets
CES-485 Approximating the Set of Pareto Optimal Solutions in Both the Decision and Objective Spaces by an Estimation of Distribution Algorithm
Get PDFMost existing multiobjective evolutionary algorithms aim at approximating the PF, the distribution of the Pareto optimal
solutions in the objective space. In many real-life applications, however, a good approximation to the PS, the distribution of the
Pareto optimal solutions in the decision space, is also required by a decision maker. This paper considers a class of MOPs, in
which the dimensionalities of the PS and PF are different so that a good approximation to the PF might not approximate the PS
very well. It proposes a probabilistic model based multiobjective evolutionary algorithm, called MMEA, for approximating the PS
and the PF simultaneously for a MOP in this class. In the modelling phase of MMEA, the population is clustered into a number
of subpopulations based on their distribution in the objective space, the PCA technique is used to detect the dimensionality of the
centroid of each subpopulation, and then a probabilistic model is built for modelling the distribution of the Pareto optimal solutions
in the decision space. Such modelling procedure could promote the population diversity in both the decision and objective spaces.
To ease the burden of setting the number of subpopulations, a dynamic strategy for periodically adjusting it has been adopted in
MMEA. The experimental comparison between MMEA and the two other methods, KP1 and Omni-Optimizer on a set of test
instances, some of which are proposed in this paper, have been made in this paper. It is clear from the experiments that MMEA
has a big advantage over the two other methods in approximating both the PS and the PF of a MOP when the PS is a nonlinear
manifold, although it might not be able to perform significantly better in the case when the PS is a linear manifold
Evolving hierarchical gene regulatory networks for morphogenetic pattern formation of swarm robots
Get PDFMorphogenesis, the biological developmental process of multicellular organisms, is a robust self-organising mechanism for pattern formation governed by gene regulatory networks (GRNs). Recent findings suggest that GRNs often show the use of frequently recurring patterns termed network motifs. Inspired by these biological studies, this paper proposes a morphogenetic approach to pattern formation for swarm robots to entrap targets based on an evolving hierarchical gene regulatory network (EH-GRN). The proposed EH-GRN consists of two layers: The upper layer is for adaptive pattern generation where the GRN model is evolved by basic network motifs, and the lower layer is responsible for driving robots to the target pattern generated by the upper layer. Obstacle information is introduced as one of environmental inputs along with that of targets in order to generate patterns adaptive to unknown environmental changes. Besides, splitting or merging of multiple patterns resulting from target movement is addressed by the inherent feature of the upper layer and the k-means clustering algorithm. Numerical simulations have been performed for scenarios containing static/moving targets and obstacles to validate the effectiveness and benefit of the proposed approach for complex shape generation in dynamic environments
A Survey on Surrogate-assisted Efficient Neural Architecture Search
Full text linkNeural architecture search (NAS) has become increasingly popular in the deep
learning community recently, mainly because it can provide an opportunity to
allow interested users without rich expertise to benefit from the success of
deep neural networks (DNNs). However, NAS is still laborious and time-consuming
because a large number of performance estimations are required during the
search process of NAS, and training DNNs is computationally intensive. To solve
the major limitation of NAS, improving the efficiency of NAS is essential in
the design of NAS. This paper begins with a brief introduction to the general
framework of NAS. Then, the methods for evaluating network candidates under the
proxy metrics are systematically discussed. This is followed by a description
of surrogate-assisted NAS, which is divided into three different categories,
namely Bayesian optimization for NAS, surrogate-assisted evolutionary
algorithms for NAS, and MOP for NAS. Finally, remaining challenges and open
research questions are discussed, and promising research topics are suggested
in this emerging field.Comment: 18 pages, 7 figure
Rethinking Individual Global Max in Cooperative Multi-Agent Reinforcement Learning
Full text linkIn cooperative multi-agent reinforcement learning, centralized training and
decentralized execution (CTDE) has achieved remarkable success. Individual
Global Max (IGM) decomposition, which is an important element of CTDE, measures
the consistency between local and joint policies. The majority of IGM-based
research focuses on how to establish this consistent relationship, but little
attention has been paid to examining IGM's potential flaws. In this work, we
reveal that the IGM condition is a lossy decomposition, and the error of lossy
decomposition will accumulated in hypernetwork-based methods. To address the
above issue, we propose to adopt an imitation learning strategy to separate the
lossy decomposition from Bellman iterations, thereby avoiding error
accumulation. The proposed strategy is theoretically proved and empirically
verified on the StarCraft Multi-Agent Challenge benchmark problem with zero
sight view. The results also confirm that the proposed method outperforms
state-of-the-art IGM-based approaches.Comment: Accept at NeurIPS 202
- …
