University of Cambridge > > Inference Group > Optimizing Similarity Functions in Various Pattern Recognition Problems

Optimizing Similarity Functions in Various Pattern Recognition Problems

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact David MacKay.

Note unusual time

Machines classify instances by using various similarity functions, such as image-based feature extractors, clustering functions, kernel functions, etc. In real world applications, however, they often work poor. Since they are characterized by a set of parameters, the choice of them hugely affects performance. To solve this parameter optimization problem, in this talk is presented optimization strategies in terms of class separability. First, a Gabor Wavelet Neural Network for facial expression recognition is presented and its optimization method is proposed. Second, an agglomerative clustering technique is presented and a strategy for choosing optimal number of clusters is proposed. A concatenation of the agglomerative clustering and Fuzzy-state Q-learning allows us to learn and predict human behavior patterns. The last part of the talk is about learning a generalized kernel function for kernel-based pattern classification. Theoretical results on a generalized kernel function and its learning via regularization are presented. Its performance is demonstrated with an application of EMG signal-based walking phase recognition.

This talk is part of the Inference Group series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2024, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity