Skip to content
Learning in Model Space based on Liquid State Machine
MATLAB C
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
RV_kernel
compute_Kernel_matrix
libsvm
lsm
prepared_datasets
GMMRV.m
LICENSE
NormalRV.m
README.md
SamplingRV.m
adj_param.m
demo.m
fisherRV.m
initpath.m
param_analysis.m

README.md

RV model-based kernel with LSM

This repository contains the source code for classification based on the liquid state machine (LSM). The paper is published in ECML-16. You can find electronic version of paper in here.

We build our project on the basis of codes provided by Fengzhen Tang (fxt126@cs.bham.ac.uk). This repository integrates work from other literatures. All the contact information is retained in the preamble. ALL necessary codes can be obtained freely from the Internet.

demo.m is a script to allow a quick view, and from which a quick start for your own project.

HOWTO

  • How to start?

    • download the data sets from website here (The data sets are not included in this repo.)
    • firstly, run initpath to include some necessary folders.
    • run NormalRV, GMMRV, fisherRV, SamplingRV to test classification.
    • we have included a data set for testing purpose. Run demo.m, you will see the experimental result on that data.
  • How to adjust reservoir size (R_no) and regression coefficient (val) to get a best fitting?

    • You may run lsm_weight_*.m directly to see the fitting error, and you may also integrate your own own algorithm in adjusting R_no and val.
    • test/test.m is a useful in finding best combination of R_no and val.
    • In addition, default settings in test/test.m should not be ignored and they are good starting points for a beginner.
  • How to adjust svm parameters in svm, i.e. cost and kp?

    • Run NormalRV, GMMRV, fisherRV, SamplingRV by feeding a list of cost and kp, the best classification accuracy will come out.

Dependencies Version

This repository includes a precompiled version of csim 1.1.1 and libsvm 3.2, including mexw64, mexa64m, mexmaci64 for usage under Windows, Linux, Mac.

The csim can be found here. N.B. the original version from the website was altered to meet the updated operational environment.

libsvm 3.2 or higher version is required. libsvm is a library for support vector machines which was obtained from here.

About

This repository is mainly contributed by Junyuan Hong(jyhong836@gmail.com), in cooperation with Yang Li.

License

This repository is distributed under the GNU General Public License.

Citation

Please cite our paper as follow, if you use our codes.

Li, Y., Hong, J., & Chen, H. (2016, September). Sequential Data Classification in the Space of Liquid State Machines. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases (pp. 313-328). Springer International Publishing.

You can’t perform that action at this time.