[Download]Energy Landscape Analysis Toolkit (ELAT) The energy landscape analysis is a powerful method for analyzing multivariate time-series data (e.g. fMRI data). We provide MATLAB code for the entire analysis described in our paper [Ezaki et al. (2017)]. RequirementsMATLAB (ver. later than R2016b) MATLAB Statistics and Machine Learning Toolbox LimitationsThe energy landscape analysis is not applicable to the data if the number of variables, N, is larger than 15.Note that estimation of the maximum entropy distribution is possible even for N > 200.The length of data is crucial. We do not encourage to use this method to very short data ( Tmax<1000). What can we do after this analysis?ELAT provides: 1. Categorization of the possible 2^N states into relevant groups based on the maximum entropy distribution. 2. Disconnectivity graph which characterizes the energy landscape. 1. can be used to label states for further analysis (e.g., elaborate a new dynamical index). 2. may characterize the difference between two groups (subject groups etc.) of data. See references listed in [Ezaki et al. (2017)] for examples of the use of the energy landscape analysis. Download ELAT Update: 1. Added the correspondences between the labeling numbers of local minima (i.e., 1,2,3,..) and state numbers shown in the console. Old versions (for archiving purpose). Updates: 1. Added accuracy of fitting, r.2. Definitions of ROI and state number were modified. Contact informationE-mail: ezaki@jamology.rcast.u-tokyo.ac.jp | How to usePreparations1. Prepare multivariate time-series data (a). (Number of variables must be smaller than 15). Each line (row) of the data file corresponds to time-series of a single variable. 2. ELAT provides a simple binarization method. If this is not suitable for your data, prepare binarized data. 3. If necessary, prepare a list of variable names (b). We provide test data files (a) "testdata.dat" and (b) "roiname.dat". Analysis4. Download ELAT and unzip it to a relevant directory. 5. Run "StartProgram.m". 6. Select data files (a) and (b). If the data (a) is already binarized, select "Binarized data"; otherwise, select "Continuous data" and set a threshold value ("0" was used in [Ezaki et al. (2017)]). 7. Select output folder where results will be saved. If you need the list of basins for further analyses, select "Save Basin List". 8. Execute. Alternatively, you can directly run this program by executing main.m.In this case, please set an appropriate path to the analyzed data file in main.m.Data structure / Definition of the state number In this case, the state number is 28 + 1 = 29. The state number ranges from 1 to 2^N (not from 0 to 2^N–1).Note: In ELAT ver. 1.0, the definition of variable names is reversed. Please use the latest version. |