RadClass

A collection of tools (including data analysis and machine learning applications) that can be used to explore radiation data (temporal and spectral) for external identification of radiation signatures.

Table of Contents

1. Installation

2. Usage

3. Data Format

Installation

Dependencies

• Python 3

Versions 3.6-3.9 are currently supported by tests. The following Python packages are required to run (found in requirements.txt):

• h5py
• numpy
• progressbar2
• matplotlib
• seaborn
• scipy
• sklearn
• hyperopt
• torch
• shadow-ssml

Modules can be imported from the repository directory (e.g. from RadClass.H0 import H0) or RadClass can be installed using pip:

python -m pip install /path/to/RadClass/. --user

Usage

RadClass is designed to be modular, allowing a user to use analysis modules as needed:

An HDF5 data file is specified as input, which is processed by DataSet. The user can specify a type of AnalysisParameters. For example, H0 for hypothesis testing, BackgroundEstimator for background estimation, etc. Processor then uses DataSet and the user specified AnalysisParameters to run, storing the results for use by the user. To see examples of how a user can initialize and run Processor, review /tests/.

Data Format

RadClass.Processor expects a data structure as follows:

The HDF5 file must have three groups:

1. A vector of epoch timestamps at which the measurement was taken.
2. If applicable, the live recording time of the detector to correct for dead time. If not applicable, a vector of 1's.
3. A data matrix of n measurements given for m bins.

Each group's name must be specified in an input dictionary: labels.

Integration occurs over the course of the data matrix.

Data rows are corrected for dead time and summed for the specified integration input length (then averaged over the integration period). If stride is specified, the working timestamp will advance forward by the specified amount. While a required input, setting stride = integration will ignore this behavior. A cache_size can be given, which will pre-slice a specified number of rows into RadClass to reduce file I/O. Otherwise, each set of integration rows will be sliced separately. A start_time and stop_time can also be specified for periods of data processing smaller than the total length of the input file. These two variables must be epoch timestamps in units of seconds, which can generally be given using the time and datetime Python modules:

time.mktime(datetime.datetime.strptime(x, "%m/%d/%Y %H:%M:%S").timetuple())

Where x is a date string. See module docs for more details.