Fast and robust fixed-point algorithms for independent component analysis

By Aapo Hyvärinen. 1999

Subfields: Statistics and Probability, Numerical Methods & Scientific Computing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistics and Probability

In Numerical Methods & Scientific Computing

Citations in the corpus

Rethinking LDA: Moment Matching for Discrete ICA (2015, neutral): “For ICA, the most notable ones are probably the FastICA algorithm [20] and the JADE algorithm [21].”
Unsupervised Feature Extraction by Time-Contrastive Learning and Nonlinear ICA (2016, methodology): “According to Corollary 1 above, after TCL we further applied linear ICA (FastICA, [15]) to the h(x;θ), and used its outputs as the final estimates of q(si).”
A Survey of Inductive Biases for Factorial Representation-Learning (2016, methodology): “The PCA components all code for global features at different 3Using the FastICA[21] algorithm”
Tensor decompositions for learning latent variable models (2012, methodology): “This method can be effectively implemented using fast descent algorithms (Hyvarinen, 1999).”
Principal Component Analysis: A Method for Determining the Essential Dynamics of Proteins (2014, methodology): “To achieve this, ICA is typically implemented using either kurtosis or an information theoretic quantity like mutual information (FastICA) as a contrast function [50].”
Whole Brain Resting-State Analysis Reveals Decreased Functional Connectivity in Major Depression (2010, methodology): “Next, the data set was decomposed into 20 sets of independent vectors which describe signal variation across the temporal (time-courses) and spatial (maps) domain by optimizing for non-Gaussian spatial source distributions using the FastICA algorithm (Hyvarinen, 1999).”
ICA-based artifact removal diminishes scan site differences in multi-center resting-state fMRI (2015, methodology): “By optimizing for non-Gaussian source estimates through a fixed-point iteration technique, we obtained component maps (Hyvärinen, 1999).”
DirectLiNGAM: A Direct Method for Learning a Linear Non-Gaussian Structural Equation Model (2011, methodology): “Given a p-dimensional random vector x and its p × n observed data matrix X, apply an ICA algorithm (FastICA using hyperbolic tangent function [12]) to obtain an estimate of A.”
Pairwise Likelihood Ratios for Estimation of Non-Gaussian Structural Equation Models (2013, methodology): “Since basic FastICA, which is an integral part of the method, has convergence problems with the basic tanh nonlinearity in the case of a small sample size, we used the stabilized version by Hyvärinen (1999) obtained in the standard FastICA package by the option " stabilize " .”
Functional segmentation of the brain cortex using high model order group PICA (2009, methodology): “Because ICA is sensitive to its initialization, this framework runs FastICA multiple times on the same data, with random initializations and clusters with a hierarchical clustering algorithm, on the results from individual runs.”
Oscillatory synchronization in large-scale cortical networks predicts perception. (2011, unknown): “To reduce remaining artifacts (e.g., small eye movements, muscle twitches, and cardiac artifacts), we applied independent component analysis (Hyvarinen, 1999; Jung et al., 2000), separately for high and low frequencies, and rejected components that reflected signal artifacts.”