Computationally E cient Transductive Machines

In this paper we propose a new algorithm for providing conndence and credibility values for predictions on a multi-class pattern recognition problem which uses Support Vector machines in its implementation. Previous algorithms which have been proposed to achieve this are very processing intensive and are only practical for small data sets. We present here a method which overcomes these limitations and can deal with larger data sets (such as the US Postal Service database). The measures of conndence and credibility given by the algorithm are shown empirically to reeect the quality of the predictions obtained by the algorithm, and are comparable to those given by the less computationally eecient method. In addition to this the overall performance of the algorithm is shown to be comparable to other techniques (such as standard Support Vector machines), which simply give at predictions and do not provide the extra conndence/credibility measures.