RAMjET: RApid MachinE-learned Triage

RAMjET is currently under development. This documentation is currently oriented toward the RAMjET development team.

RAMjET is a framework for producing neural networks to characterize phenomena in astrophysical photometric data.

The basic idea of RAMjET is to take photometric data (either in the form of lightcurves or time-series of flux frames) and automatically search for patterns in this data which correspond to astrophysical phenomena. The pipeline will then automatically characterized these discovered patterns. These characterizations can come in several forms. The simplest case would be classifying a lightcurve as containing a specific phenomena (e.g., this lightcurve contains a microlensing event or not). However, the characterization could be more complex, such as providing a classification for each time step (e.g., this time step contains a microlensing event) or even providing quantitative characterizations (e.g., this time step is being magnified by a factor of X due to a microlensing event).

To produce such a neural network pipeline which can produce good predictions two main components need to be handled. First, the data for training and testing the network needs to be prepared. Second, the network architecture along with an appropriate training system needs to be designed. The main development focus of this project is provide a framework to easily complete these two tasks, and to demonstrate the framework on specific examples.

Understanding RAMjET

• Quick start tutorial
  • Tutorial Summary
  • Install
  • Download the data
  • Train the network
  • Using the trained network to make predictions
• Simple TESS data interface tutorial
  • Tutorial Summary
  • Setup
  • Download a TESS lightcurve
  • Plot a lightcurve from MAST
  • Checking for known planets
  • Checking for known variability
• Input data types
• Output data types
• Using the Python API for TESS data
  • Listing what targets have data
  • Finding observation data products
  • Downloading the data products