The NOBIAS algorithm (NOnparametric Bayesian Inference for Anomalous diffusion in Single-molecule tracking) is a two-module algorithm for analysis of multi-diffusive state SPT datasets that predicts the anomalous diffuion type for each state.
Written by Ziyuan Chen at the University of Michigan.
Download the entire folder and unzip if you downloaded the .zip folder. Change the working directory in Matlab to this folder and call the functions in the Matlab command window as described in the User Guide.
NOBIAS also requires the lightspeed toolbox in Matlab which can be found at https://github.com/tminka/lightspeed
Running the RNN module needs the MATLAB Deep Learning toolbox: https://www.mathworks.com/products/deep-learning.html
Running the RNN module needs the MATLAB Mapping toolbox: https://www.mathworks.com/products/mapping.html
Use the quick start guide for now: https://github.com/BiteenMatlab/NOBIAS/blob/main/NOBIAS%20Quick%20Start%20Guide.pdf
A detailed user guide will be finished very soon.
Great thanks to the work of Dr. Emily B. Fox who inspired me to use her sticky HDP-HMM for SPT datasets.
randdirichlet and randiwishart by Emily B. Fox and Erik B. Sudderth.(ebfox[at]alum[dot]mit[dot]edu and sudderth[at]cs[dot]brown[dot]edu)
Copyright (C) 2009, Emily B. Fox and Erik B. Sudderth.
The HDPHMM module majorly adapts the algorithm in:
An HDP-HMM for Systems with State Persistence E. B. Fox, E. B. Sudderth, M. I. Jordan, and A. S. Willsky Proc. Int. Conf. on Machine Learning, July, 2008.
lightspeed toolbox MATLAB, Tom Minka.
Copyright (c) 2017 Tom Minka
The RNN module uses the LSTM arichtecture of the RANDI python package and implement in MATLAB. Argun, A., Volpe, G., and Bo, S. (2021). Classification, inference and segmentation of anomalous diffusion with recurrent neural networks. J. Phys. A: Math. Theor. doi:10.1088/1751-8121/ac070a.
GNU GENERAL PUBLIC LICENSE
Version 3, 29 June 2007
See LICENSE.txt