PLVM - Probabilistic Latent Variable Modeling toolbox for multisubject data
The Probabilistic Latent Variable Modeling Toolbox for Multisubject Data holds a collection of latent variable algorithms implemented in Matlab™. The algorithms support the use of graphical processing units (GPUs) for high performance computing. All code can be used freely in research and other non-profit applications. If you publish results obtained with this toolbox we kindly ask that our and other relevant sources are properly cited.
This toolbox has been developed at:
The Technical University of Denmark, Department for Applied Mathematics and Computer Science, Section for Cognitive Systems.
The toolbox was developed in connection with the Brain Connectivity project at DTU (https://brainconnectivity.compute.dtu.dk/) .
- Probabilistic Sparse Factor Analysis (psFA). Models subject specific heteroscedastic feature/voxel noise.
- Probabilistic Sparse Principal Component Analysis (psPCA). Models subject specific homoscedastic noise.
Common algorithm properties
- No parameters need to be set by default.
- All models support modeling multiple subjects.
- An approximate model solution is found using variational inference.
- The evidence lowerbound is calculated (an approximation to log likelihood).
- Estimating number of components using Automatic Relevance Determination (ARD).
- Ability to individually turn off modeling aspects such as sparsity, ARD and noise modeling.
- Demostration on a toy example showing model settings and their effect
- Synthetic experiments in the original psFA article (which is currently under review).