Automated document retrieval and classification is of central importance in many contexts; our main motivating goal is the efficient classification and retrieval of "interests" on the internet when only positive information is available. In this paper, we show how a simple feed-forward neural network can be trained to filter documents under these conditions, and that this method seems to be superior to modified methods (modified to use only positive examples), such as Rocchio, Nearest Neighbor, Naive-Bayes, Distance-based Probability and One-Class SVM algorithms. A novel experimental finding is that retrieval is enhanced substantially in this context by carrying out a certain kind of uniform transformation ("Hadamard") of the information prior to the training of the network.
Bibliographical noteFunding Information:
This work was partially supported by HIACS, the Haifa Interdisciplinary Center for Advanced Computer Science. This work forms part of the doctoral thesis of the second author who was supported by a University of Haifa fellowship during his studies, and afterwards hosted by the Neurocomputation Laboratory situated in the Caesarea Rothschild Institute for Interdisciplinary Computer Science. We thank Nathalie Japkowicz for lending us some Matlab software which was used during a revision of this paper. The first author thanks Oxford University for its hospitality during his sabbatical visit.
This work was partially supported by HIACS, the Haifa University Interdisciplinary Center for Advanced Computer Science, and the Neurocomputation Laboratory located at the Caesarea Rothschild Institute for Interdisciplinary Computer Science.
- Automated document retrieval
- Bottleneck neural network
- Feed-forward neural networks
- Machine learning
- One-class classification
ASJC Scopus subject areas
- Computer Science Applications
- Cognitive Neuroscience
- Artificial Intelligence