NEMS is the Neural Encoding Model System, a set of tools for fitting computational models for the neural encoding of sensory stimuli. Written in Python, migrated from a Matlab tool designed to do something similar (NARF).
NEMS_WEB is a GUI that accompanies the core NEMS software in the form of a web server. NEMS must be installed in order for NEMS_WEB to function.
General Overview:
-
NEMS models are described by keyword strings, e.g. 'fb18ch100_wc03_fir10_dexp_fit01_nested10'. These keywords tell the fitter how to load the data, which model functions to include in the model, and how to fit the model.
-
The NEMS models are essentially a modified stack, with each level of the stack being the output of a function applied to the previous element of the stack.
-
Models can be fit locally using the core NEMS software, through the NEMS_WEB interface at neuralprediction.org, or through a locally-hosted instance of the NEMS_WEB server (typically at localhost:8000).
Ongoing: expand this information in NEMS Wiki
- TODO
-
Dependencies
- NEMS package available at https://bitbucket.org/lbhb/nems and its dependencies.
- Additional NEMS_WEB-specific dependencies: pandas, flask, mpld3, bokeh, flask-socketio, eventlet, flask-login, flask-WTF, bcrypt, boto3, seaborn, gevent flask-assets
-
Demos
- A sample database will be downloaded on first launch if no database information is provided. This sample contains a limited amount of fitting results that can be used to explore the NEMS_WEB interface. A sample data file that corresponds to the first cell and first model in the selection lists can be used to fit models.
- LBHB team