scientific article; zbMATH DE number 1302403
The purpose of this chapter is to present results for several related learning models, which may be useful both in machine learning and in the analysis of human learning. One motivation has been our desire to develop efficient and powerful methods of concept learning to use in our work on machine learning of natural language. In this earlier work, concepts were assumed known and the learning concentrated on language. It is obviously important to combine both concept and language learning for many kinds of situations. In unpublished research we have already used the learning models proposed here in the early stages of robotic concept and language learning. The models proposed are compared in two major ways: their comparative rates of concept learning, and their comparative transfer performance when the learning of one concept is followed by another. The several models are compared on six sets of data. The first section that follows presents the six models and related theory, beginning with a multivariate normal learning model. The second section introduces a model for comparatively evaluating the performance of the several learning models. The third presents and analyzes the learning results for the several models of data sets. The fourth section is devoted to problems of transfer, and to a particular data set. The final section compares the models and results to a variety of other models discussed in the literature. For the entire collection see [Zbl 0912.00034].
- Knowledge transfer in SVM and neural networks
- Learning feed-forward multi-nets
- The learning problem of multi-layer neural networks
- Machine learning in vector models of neural networks
- Learning and generalization in multineuron interacting feed-forward neural networks
- A theoretical framework for deep transfer learning
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4247708)
