Neural network module

DEERNet is a collection of functions that make and use deep neural networks for processing DEER data. The approach is described in detail in (https://dx.doi.org/10.1126/sciadv.aat5218). The reasons why it works so well are analysed in (https://arxiv.org/abs/1912.01498).

Using DEERNet

Import your data with elexsys2deernet.m and feed it into deernet.m - examples are provided in examples/deernet directory. You will get outputs that looks like the following:

Functions

deer_lib_gen.m

Generates a library of simulated DEER data for use in neural network training and validation.

deernet.m

Uses a curated ensemble of neural networks to extract the distance distribution from DEER data.

deernet_wrapper.m

Compilable DEERNet wrapper for use outside Matlab and Spinach.

descramble.m

Weight matrix descrambling using Tikhonov smoothness criterion.

dist_range.m

Distance range estimation for a given time grid.

elexsys2deernet.m

Prepares standard Bruker Elexsys datasets for input into the deernet.m function.

left_diag.m

Weight matrix descrambling using maximum diagonality criterion.

process_using.m

Runs DEER data processing using a specified neural network file.

train_one_net.m

Trains a single neural network using supplied parameters.

Service functions

quality_control.m

Internal heuristics designed to catch malformed inputs.

Training your own networks

Training your own networks will require powerful hardware - IK's workstation has 32 Xeon cores, 1 TB of RAM, and three NVidia Titan V cards. Network sets included with the public versions of DEERNet took weeks to train.

1. Go into experiments/deernet and edit library_params.m file.
2. Delete all *.mat files from network directories.
3. Edit training_script.m file to match your hardware configuration.
4. Launch training_script.m file and hope that your computer holds up.

Version 2.5, authors: Ilya Kuprov, Steve Worswick, Jake Amey, Jake Keeley, Tajwar Choudhury