Pattern Recognition and Machine Learning
Citations
238 citations
Cites background from "Pattern Recognition and Machine Lea..."
...the bigram feature coefficients) that needed to be estimated from training data, which can potentially increase generalization error arising from increased model complexity [45]....
[...]
238 citations
Cites methods from "Pattern Recognition and Machine Lea..."
...A plethora of regression methods have been employed to tackle the above mentioned problems including linear and ridge [4], Support Vector [31], Boosted [13], Gaussian process [26], and more recently, Deep Neural Nets [18]....
[...]
237 citations
Cites methods from "Pattern Recognition and Machine Lea..."
...Based on the above, we utilize the Gamma distribution for the hyperparameters , and , since it is the conjugate prior for the inverse variance (precision) of the Gaussian distribution....
[...]
237 citations
Cites background or methods from "Pattern Recognition and Machine Lea..."
...Each factor is a set of 1-D MRFs whose exact marginals are calculated using the Baum-Welch algorithm [3]....
[...]
...PLSA can be viewed as a probabilistic generalization of PCA – a low-rank nonnegative approximation of the matrix of (empirical word probabilities for each document)× (documents) using factor matrices {p(w|t)}×{p(t|d)} obtained by Expectation-Maximization (EM) [3]....
[...]
...For an introduction to the latter three methods see [3], and for more details on EP and its relation to loopy BP, variational and other approximations see [16]....
[...]
...However there are various methods for approximating marginals, including (structured) Gibbs sampling [8], Variational Mean-Field (VMF), Loopy Belief Propagation (LBP), and Expectation Propagation (EP)....
[...]
...When the node marginals are multi-modal, EP tends to smooth over the peaks and hence over-estimate the marginal entropies, but in practice its marginals are often found to be more accurate than those of VMF and LBP [3]....
[...]
237 citations
Cites methods from "Pattern Recognition and Machine Lea..."
...For this fully connected graph, we can re-order its nodes as follows [3]:...
[...]