arXiv subjects

David Howard

Publications and source records attributed to David Howard.

61 records · Page 4Linked to original sources

arXiv2017

Differential Evolution and Bayesian Optimisation for Hyper-Parameter Selection in Mixed-Signal Neuromorphic Circuits Applied to UAV Obstacle Avoidance

The Lobula Giant Movement Detector (LGMD) is a an identified neuron of the locust that detects looming objects and triggers its escape responses. Understanding the neural principles and networks that lead to these fast and robust responses can lead to the design of efficient facilitate obstacle avoidance strategies in robotic applications. Here we present a neuromorphic spiking neural network model of the LGMD driven by the output of a neuromorphic Dynamic Vision Sensor (DVS), which has been optimised to produce robust and reliable responses in the face of the constraints and variability of its mixed signal analogue-digital circuits. As this LGMD model has many parameters, we use the Differential Evolution (DE) algorithm to optimise its parameter space. We also investigate the use of Self-Adaptive Differential Evolution (SADE) which has been shown to ameliorate the difficulties of finding appropriate input parameters for DE. We explore the use of two biological mechanisms: synaptic plasticity and membrane adaptivity in the LGMD. We apply DE and SADE to find parameters best suited for an obstacle avoidance system on an unmanned aerial vehicle (UAV), and show how it outperforms state-of-the-art Bayesian optimisation used for comparison.

cs.NE↗

arXiv2016

A rainbow $r$-partite version of the Erdős-Ko-Rado theorem

Let $f(n,r,k)$ be the minimal number such that every hypergraph larger than $f(n,r,k)$ contained in $\binom{[n]}{r}$ contains a matching of size $k$, and let $g(n,r,k)$ be the minimal number such that every hypergraph larger than $g(n,r,k)$ contained in the $r$-partite $r$-graph $[n]^{r}$ contains a matching of size $k$. The Erdős-Ko-Rado theorem states that $f(n,r,2)=\binom{n-1}{r-1}$~~($r \le \frac{n}{2}$) and it is easy to show that $g(n,r,k)=(k-1)n^{r-1}$. The conjecture inspiring this paper is that if $F_1,F_2,\ldots,F_k\subseteq \binom{[n]}{r}$ are of size larger than $f(n,r,k)$ or $F_1,F_2,\ldots,F_k\subseteq [n]^{r}$ are of size larger than $g(n,r,k)$ then there exists a rainbow matching, i.e. a choice of disjoint edges $f_i \in F_i$. In this paper we deal mainly with the second part of the conjecture, and prove it for $r\le 3$. \vspace{.1cm} We also prove that for every $r$ and $k$ there exists $n_0=n_0(r,k)$ such that the $r$-partite version of the conjecture is true for $n>n_0$.

math.CO↗

arXiv2016

Cross-intersecting pairs of hypergraphs

Two hypergraphs $H_1,\ H_2$ are called {\em cross-intersecting} if $e_1 \cap e_2 \neq \emptyset$ for every pair of edges $e_1 \in H_1,~e_2 \in H_2$. Each of the hypergraphs is then said to {\em block} the other. Given parameters $n,r,m$ we determine the maximal size of a sub-hypergraph of $[n]^r$ (meaning that it is $r$-partite, with all sides of size $n$) for which there exists a blocking sub-hypergraph of $[n]^r$ of size $m$. The answer involves a fractal-like (that is, self-similar) sequence, first studied by Knuth. We also study the same question with $\binom{n}{r}$ replacing $[n]^r$.

math.CO↗

arXiv2015

Evolving Unipolar Memristor Spiking Neural Networks

Neuromorphic computing --- brainlike computing in hardware --- typically requires myriad CMOS spiking neurons interconnected by a dense mesh of nanoscale plastic synapses. Memristors are frequently citepd as strong synapse candidates due to their statefulness and potential for low-power implementations. To date, plentiful research has focused on the bipolar memristor synapse, which is capable of incremental weight alterations and can provide adaptive self-organisation under a Hebbian learning scheme. In this paper we consider the Unipolar memristor synapse --- a device capable of non-Hebbian switching between only two states (conductive and resistive) through application of a suitable input voltage --- and discuss its suitability for neuromorphic systems. A self-adaptive evolutionary process is used to autonomously find highly fit network configurations. Experimentation on a two robotics tasks shows that unipolar memristor networks evolve task-solving controllers faster than both bipolar memristor networks and networks containing constant nonplastic connections whilst performing at least comparably.

cs.NE↗

arXiv2015

A Cognitive Architecture Based on a Learning Classifier System with Spiking Classifiers

Learning Classifier Systems (LCS) are population-based reinforcement learners that were originally designed to model various cognitive phenomena. This paper presents an explicitly cognitive LCS by using spiking neural networks as classifiers, providing each classifier with a measure of temporal dynamism. We employ a constructivist model of growth of both neurons and synaptic connections, which permits a Genetic Algorithm (GA) to automatically evolve sufficiently-complex neural structures. The spiking classifiers are coupled with a temporally-sensitive reinforcement learning algorithm, which allows the system to perform temporal state decomposition by appropriately rewarding "macro-actions," created by chaining together multiple atomic actions. The combination of temporal reinforcement learning and neural information processing is shown to outperform benchmark neural classifier systems, and successfully solve a robotic navigation task.

cs.NE↗

arXiv2013

On a Generalization of the Ryser-Brualdi-Stein Conjecture

A rainbow matching for (not necessarily distinct) sets F_1,...,F_k of hypergraph edges is a matching consisting of k edges, one from each F_i. The aim of the paper is twofold - to put order in the multitude of conjectures that relate to this concept (some of them first presented here), and to present some partial results on one of these conjectures, that seems central among them.

math.CO↗

arXiv2012

Revolutionaries and Spies

Let $G = (V,E)$ be a graph and let $r,s,k$ be positive integers. "Revolutionaries and Spies", denoted $\cG(G,r,s,k)$, is the following two-player game. The sets of positions for player 1 and player 2 are $V^r$ and $V^s$ respectively. Each coordinate in $p \in V^r$ gives the location of a "revolutionary" in $G$. Similarly player 2 controls $s$ "spies". We say $u, u' \in V(G)^n$ are adjacent, $u \sim u'$, if for all $1 \leq i \leq n$, $u_i = u'_i$ or ${u_i,u'_i} \in E(G)$. In round 0 player 1 picks $p_0 \in V^r$ and then player 2 picks $q_0 \in V^s$. In each round $i \geq 1$ player 1 moves to $p_i \sim p_{i-1}$ and then player 2 moves to $q_i \sim q_{i-1}$. Player 1 wins the game if he can place $k$ revolutionaries on a vertex $v$ in such a way that player 1 cannot place a spy on $v$ in his following move. Player 2 wins the game if he can prevent this outcome. Let $s(G,r,k)$ be the minimum $s$ such that player 2 can win $\cG(G,r,s,k)$. We show that for $d \geq 2$, $s(\Z^d,r,2)\geq 6 \lfloor \frac{r}{8} \rfloor$. Here $a,b \in \Z^{d}$ with $a \neq b$ are connected by an edge if and only if $|a_i - b_i| \leq 1$ for all $i$ with $1 \leq i \leq d$.

math.CO↗