Installation

JRCLUST installation guide

Requirements

1. Matlab (R2014b or newer) and toolboxes: Parallel computing, Signal processing, Statistics and Machine Learning, Image processing toolbox.
2. RAM size should be greater than 1/4 or the recording file size.
3. NVidia GPU cards (Compute version 3.0 or higher, which includes Kepler, Maxwell, Pascal-based GPUs). Note that Pascal-based GPU is only supported by CUDA 8.0 or higher, which requires Matlab 2017a or higher.
4. NVIDIA CUDA toolkit (Install a correct version supported by your Matlab version) R2017a: CUDA 8.0, R2016b: CUDA 7.5, R2016a: CUDA 7.5, R2015b: CUDA 7.0, R2015a: CUDA 6.5, R2014b: CUDA 6.0
5. Latest NVidia GPU driver http://www.nvidia.com/Download/index.aspx\
6. Microsoft Microsoft Visual Studio 2013 Express. Download from here.

Installing prerequisites

1. Install Microsoft Visual Studio 2012 (Express edition, free) or 2013 (Community edition, free) https://www.visualstudio.com/en-us/news/releasenotes/vs2013-community-vs
2. Set the system path to the Microsoft Visual C compiler (cl.exe) by editing "PATH" variable in "Environment Variables" in Windows. for Visual studio 2012, "cl.exe" path is "C:\Program Files (x86)\Microsoft Visual Studio 11.0\VC\bin\amd64" Visual studio 2013 is recommend for matlab version R2016a or newer. Visual studio 2012 is recommend for matlab version R2015b or older.
3. Add "C:\Windows\System32" to the system path (required for Kilosort extension).
4. Install a correct version of NVIDIA CUDA toolkit that is supported by Matlab. To check the CUDA version supprted by Matlab, run "gpuDevice" in Matlab and check "ToolkitVersion". R2017a: CUDA 8.0, R2016b: CUDA 7.5, R2016a: CUDA 7.5, R2015b: CUDA 7.0, R2015a: CUDA 6.5, R2014b: CUDA 6.0 The below is the known relationship between Matlab and CUDA versions (links provided below)
   CUDA 8.0 (current) link: https://developer.nvidia.com/cuda-downloads CUDA 7.5 link: https://developer.nvidia.com/cuda-75-downloads-archive CUDA 7.0 link: https://developer.nvidia.com/cuda-toolkit-70 CUDA 6.5 link: https://developer.nvidia.com/cuda-toolkit-65 CUDA 6.0 link: https://developer.nvidia.com/cuda-toolkit-60
5. Add NVidia C-compiler (nvcc.exe) to the system path. For CUDA 8.0, the path is "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\bin" For CUDA 7.5, the path is "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5\bin" For CUDA 7.0, the path is "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.0\bin" For CUDA 6.5, the path is "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v6.5\bin"
6. Run Nsight Monitor > Nsight Monitor options (bottom right) > set "WDDM TDR enabled" to false.

Installation from the zip file

1. Download jrc3.zip from www.jrclust.org and unzip (or the latest jrc version).
2. Start Matlab and "cd" to the unzipped location(e.g. "C:\jrc3").
3. Run "jrc install".
4. To update, you need to download jrc3.zip from www.jrclust.org (or the latest version).

Test using the included sample data

1. Download sample data by running

jrc download sample

2. Create a new parameter file

jrc makeprm sample.bin sample.prb This creates "sample_sample.prm" file.

3. Display the raw traces

jrc traces sample_sample.prm or you can also run jrc traces You can omit sample_sample.prm since JRCLUST remembers the current parameter file. Run "jrc clear" to clear the JRCLUST memory.

4. Run spike sorting

jrc spikesort

5. Manual verification/sorting

jrc manual

6. You can clear the memory and delete all output files for a given parameter file by running

jrc clear sample_sample.prm

7. To test the full suite, run

jrc unit-test