This is all the machine learning code I've used for Nab and Ca45 data. If you want to just jump in, check out
- DNP_2019 for the most recent code
- data-exploration for all my data-exploration code
- others for more side problems (optimized synthetic data generation, optimal proprocessing filtering, etc)
Since this is a comprehensive repository of all the machine learning I have done with Ca45/Nab data (the data is very similar), there are several different purposes all togther:
- data exploration - using real data, try to find what is in it. the goals here are
- what is in the data?
- what are fast/accurate ways to select different features?
- pileup identification - this answers several questions:
- what is the optimal preprocessing for training ML to recognize two events in the same readout (pileup)?
- what architecture performs best?
- what is the best structure for the output?
- how can i compare different methods?
Many of these use a few similar things:
- synthetic data generators
- traditional signal processing methods
- data wrangling - converting from a custom binary data type, chopping off uninformative data, etc
- preprocessing -- normalizing, making cuts to the data, etc
- keras/sklearn toolkits