A fast fixed-point algorithm for independent component analysis
Neural Computation
Morphological bidirectional associative memories
Neural Networks
Morphological associative memories
IEEE Transactions on Neural Networks
FL-GrCCA: A granular computing classification algorithm based on fuzzy lattices
Computers & Mathematics with Applications
Hi-index | 0.00 |
Pursuing an analogy to the Independent Component Analysis (ICA) we propose a Lattice Independent Component Analysis (LICA), where ICA signal sources correspond to the so-called endmembers and the mixing matrix corresponds to the abundance images. We introduce an approach to fMRI analysis based on a Lattice Computing based algorithm that induces endmembers from the data. The endmembers obtained this way are used to compute the linear unmixing of each voxel's time series independently. The resulting mixing coefficients roughly correspond to the General Linear Model (GLM) estimated regression parameters, while the set of endmembers corresponds to the GLM design matrix. The proposed approach is model free in the sense that the design matrix is not fixed a priori but induced from the data. Our approach does not impose any assumption on the probability distribution of the data. We show on a well known case study that this unsupervised approach discovered activation patterns are similar to the ones detected by an Independent Component Analysis (ICA).