Aapo Hyvärinen

Menu

New:

NeurIPS2021: Shared Independent Component Analysis for Multi-Subject Neuroimaging

NeurIPS2021: Disentangling Identifiable Features from Noisy Data with Structured Nonlinear ICA

AISTATS2021: Causal Autoregressive Flows

AISTATS2021: Independent innovation analysis for nonlinear vector autoregressive process

JMLR: Information criteria for non-normalized models

JNE: Uncovering the Structure of Clinical EEG Signals with Self-supervised Learning

Software

[Mostly programmed by collaborators, but implementing algorithms I have (co-)developed]

Deep unsupervised learning

Structured Nonlinear ICA using JAX

ICE-BEeM using Pytorch

iVAE using PyTorch

Independent Innovation Analysis using PyTorch

Nonlinear ICA using HMM using JAX

Time-Contrastive Learning using TensorFlow 1 (a bit old)

Linear unsupervised learning

FastICA: Fast Independent Component Analysis (for Matlab, see scikit-learn for a good Python version)

ICASSO: Analyzing reliability of independent components (for Matlab, see here for Python)

LiNGAM: Causal discovery based on non-Gaussianity (for various systems: Python, R, Matlab)

Shared Independent Component Analysis for multi-view data (in Python)

Natural image statistics / visual modelling (for Matlab)

Natural Image Statistics package (code for the book);
alternatively the imageica package

Neuroimaging data analysis (for Matlab)

SpeDeBox: Decoding EEG/MEG using spectral infomation

OCF: Analysing variability (nonstationarity) of connectivity

ISCTEST: Testing independent components

Fourier-ICA: Improved ICA by time-frequency transforms