Pattern Recognition and Machine Learning
Citations
363 citations
Cites background from "Pattern Recognition and Machine Lea..."
...It follows that the posterior distribution for W, which is proportional to the product of the prior and the likelihood function [6], is given by:...
[...]
362 citations
Cites background from "Pattern Recognition and Machine Lea..."
...The SVM is considered the most powerful and favorable classifier in the statistical learning community (Alpaydin, 2004; Bishop, 2006; Cherkassky & Mulier, 2007)....
[...]
359 citations
Cites background or methods from "Pattern Recognition and Machine Lea..."
...Using all the features for machine learning might deteriorate recognition performance due to the curse of dimensionality (Bishop, 2006)....
[...]
...It is also interesting to see that in these tests, using six LDA transformed features delivers even higher accuracy than using dozens of SFS features, indicating that the effective reduction of feature dimensionality offered by LDA indeed contributes to recognition performance....
[...]
...Similar to the case where spectral features are evaluated individually, the MSFs achieve the highest overall accuracy when combined with prosodic features, and up to 91.6% recognition rate can be obtained using LDA with SN. Applying LDA with SN also gives the best recognition performance for icates the best performance in each test....
[...]
...The numeric results of applying SFS and LDA techniques to the FDR screened feature pools are detailed in Table 2 with SVMs employed for classification and the results averaged over the 10 cross-validation trials....
[...]
...Since the maximum rank of Sb for a C-class problem is C 1, the maximum number of LDA features is also C 1....
[...]
356 citations
355 citations
Cites background or methods from "Pattern Recognition and Machine Lea..."
...An explanation of the reasoning involved in obtaining the algorithm is beyond the scope of this paper but may be found in (Bishop, 2006, Chapter 8)....
[...]
...If observations do not distinguish between different concept-values one would expect the messages for those values to be the same....
[...]
...The technique is based on the loopy belief propagation (LBP) algorithm (Bishop, 2006, Chapter 8)....
[...]