English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

We investigate solving discrete optimisation problems using the estimation ofdistribution (EDA) approach via a novel combination of deep beliefnetworks(DBN) and inductive logic programming (ILP).While DBNs are used tolearn the structure of successively better feasible solutions,ILP enables theincorporation of domain-based background knowledge related to the goodness ofsolutions.Recent work showed that ILP could be an effective way to use domainknowledge in an EDA scenario.However,in a purely ILP-based EDA,samplingsuccessive populations is either inefficient or not straightforward.In ourNeuro-symbolic EDA,an ILP engine is used to construct a model for goodsolutions using domain-based background knowledge.These rules are introduced asBoolean features in the last hidden layer of DBNs used for EDA-basedoptimization.This incorporation of logical ILP features requires some changeswhile training and sampling from DBNs: (a)our DBNs need to be trained with datafor units at the input layer as well as some units in an otherwise hiddenlayer, and (b)we would like the samples generated to be drawn from instancesentailed by the logical model.We demonstrate the viability of our approach oninstances of two optimisation problems: predicting optimal depth-of-win for theKRK endgame,and jobshop scheduling.Our results are promising: (i)On eachiteration of distribution estimation,samples obtained with an ILP-assisted DBNhave a substantially greater proportion of good solutions than samplesgenerated using a DBN without ILP features, and (ii)On termination ofdistribution estimation,samples obtained using an ILP-assisted DBN contain morenear-optimal samples than samples from a DBN without ILP features.These resultssuggest that the use of ILP-constructed theories could be useful forincorporating complex domain-knowledge into deep models for estimation ofdistribution based procedures.

Neuro-symbolic EDA-based Optimisation using ILP-enhanced DBNs