<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-0702">
  <Title>Combining a self-organising map with memory-based learning</Title>
  <Section position="4" start_page="0" end_page="0" type="intro">
    <SectionTitle>
INPUTS
MAP
</SectionTitle>
    <Paragraph position="0"> units respond to the inputs. The map unit whose weight vector is closest to the input vector becomes the winner. During training, after a winner is chosen, the weight vectors of the winner and a neighbourhood of surrounding units are nudged towards the current input.</Paragraph>
    <Paragraph position="1"> During training, the weight vectors of winning unit and a set of units within the neighbourhood of the winner are nudged, by an amount determined by the learning rate, towards the input vector. Over time the size of the neighbourhood is decreased. Sometimes the learning rate may be too. At the end of training, the units form a map of the input space that reflects how the input space was sampled in the training data. In particular areas of input space in which there were a lot of inputs will be mapped in finer detail, using more units than areas where the inputs were sparsely distributed.</Paragraph>
    <Paragraph position="2"> 3 Why use a SOM for memory editing The SOM was chosen because the input is matched to the unit with the closest weight vector. Thus it is motivated by the same principle as used to find the closest match in MBL. It is thus hoped that the SOM can minimise the risk of failing to select the closest match, since the subset will be chosen according to similarity.</Paragraph>
    <Paragraph position="3"> However, Daelemans et al (1999b) claim that, in language learning, pruning the training items is harmful. When they removed items from the memory base on the basis of their typicality (i.e. the extent to which the items were representative of other items belonging to the same class) or their class prediction strength (i.e. the extent to which the item formed a predictor for its class), the generalisation performance of the MBL system dropped across a range of language learning tasks.</Paragraph>
    <Paragraph position="4"> The memory editing approach used by Daelemans et al removes training items independently of the novel items, and the remainder are used for matching with all novel items. If one selects a different subset for each novel item based on similarity to the novel item, then maybe the risk of degrading performance in memory editing will be reduced. This work aims to achieve precisely this.</Paragraph>
  </Section>
class="xml-element"></Paper>