Mixtures of probabilistic principal component analyzers
Citations
10,141 citations
3,441 citations
3,252 citations
Cites background or methods from "Mixtures of probabilistic principal..."
...In order to reduce computational load, therefore, we assumed a block-diagonal form of the data covariance matrix for the initial PCA dimensionality reduction, which is part of the spatial PICA decomposition....
[...]
...Keywords: functional magnetic resonance imaging; brain connectivity; resting-state fluctuations; independent component analysis...
[...]
...If we assume that the source distributions p(s) are Gaussian, the model then reduces to probabilistic principal component analysis (PCA) (Tipping & Bishop 1999) and we can use Bayesian model selection criteria....
[...]
...Probabilistic PCA is used to infer upon the unknown number of sources and results in an estimate of the noise and a set of spatially whitened observations....
[...]
...The spatial maps obtained from a PCA decomposition (figure 2c) have w0 spatial correlation, and fail to identify the ‘true’ spatial maps....
[...]
2,597 citations
Cites methods from "Mixtures of probabilistic principal..."
...If we assume that the source distributions are Gaussian, the probabilistic ICA model (2) reduces to the probabilistic PCA model [20]....
[...]
...At the first stage we employ probabilistic PCA (PPCA, [20]) in order to find an appropriate linear subspace which contains the sources....
[...]
2,298 citations
References
19,056 citations
17,446 citations
13,033 citations
"Mixtures of probabilistic principal..." refers background in this paper
...(4.5) Thus the updates for π̃i and µ̃i correspond exactly to those of a standard gaussian mixture formulation (e.g., see Bishop, 1995)....
[...]
...This can be achieved with the use of a Lagrange multiplier λ (see Bishop, 1995) and maximizing 〈LC〉 + λ ( M∑ i=1 πi − 1 ) ....
[...]
...Examples include principal curves (Hastie & Stuetzle, 1989; Tibshirani, 1992), multilayer autoassociative neural networks (Kramer, 1991), the kernel-function approach of Webb (1996), and the generative topographic mapping (GTM) of Bishop, Svensén, and Williams (1998). An alternative paradigm to such global nonlinear approaches is to model nonlinear structure with a collection, or mixture, of local linear submodels....
[...]
10,656 citations
"Mixtures of probabilistic principal..." refers background in this paper
...A complementary property of PCA, and that most closely related to the original discussions of Pearson (1901) , is that the projection onto the principal subspace minimizes the squared reconstruction error P ktn i^tnk2....
[...]