A view to a KILL: Knowledge Injection via Lambda Layer
by Matteo Magnini, Giovanni Ciatto, and Andrea Omicini
We propose KILL (Knowledge Injection via Lambda Layer) as a novel method for the injection of symbolic knowledge into neural networks (NN) allowing data scientists to control what the network should (not) learn. Unlike other similar approaches, our method does not (i) require ground input formulae, (ii) impose any constraint on the NN undergoing injection, (iii) affect the loss function of the NN. Instead, it acts directly at the backpropagation level, by increasing penalty whenever the NN output is violating the injected knowledge. Experiments are reported to demonstrate the potential (and limits) of our approach.
@incollection{kill-woa2022,
author = {Magnini, Matteo and Ciatto, Giovanni and Omicini, Andrea},
booktitle = {WOA 2022 -- 23rd Workshop ``From Objects to Agents''},
dblp = {conf/woa/MagniniCO22},
editor = {Ferrando, Angelo and Mascardi, Viviana},
iris = {11585/899373},
issn = {1613-0073},
keywords = {symbolic knowledge injection; AI; ML; neural networks; KILL; PSyKI},
month = nov,
numpages = 16,
pages = {61--76},
publisher = {Sun SITE Central Europe, RWTH Aachen University},
scholar = {514437633827732665},
scopus = {2-s2.0-85142481549},
series = {CEUR Workshop Proceedings},
subseries = {AIxIA Series},
title = {A view to a {KILL}: Knowledge Injection via Lambda Layer},
url = {http://ceur-ws.org/Vol-3261/paper5.pdf},
urlopenaccess = {http://ceur-ws.org/Vol-3261/paper5.pdf},
urlpdf = {http://ceur-ws.org/Vol-3261/paper5.pdf},
volume = 3261,
year = 2022
}