Asimov Use Cases and Mapping

This document explains how to construct a YAML production specification for asimov and how those values map to the RIFT pipeline.

User-Facing Use Cases

RIFT is integrated into asimov to support several primary analysis workflows:

  1. Production-style PE: Full-scale parameter estimation runs.

  2. PP Plots: Posterior predictive checks for consistency.

  3. Single-event Analysis: Deep-dives into individual gravitational wave events.

  4. Population Analyses: Statistical studies across multiple events.

Writing a Template YAML Specification

A production specification in the asimov ledger defines the “what” and “how” of a RIFT run. Below is a simplified example of a production entry:

- Prod_RIFT_01:
    pipeline: rift
    waveform:
       approximant: IMRPhenomPv3
       maximum mode: 4
    likelihood:
       assume:
          precessing: True
          eccentric: True
       start frequency: 20
    sampler:
       cip:
          sampling method: "AV"
          explode jobs: 3
       ile:
          n eff: 10
    priors:
       spin 1:
          maximum: 0.99
       luminosity distance:
          maximum: 5000
    needs:
      - Prod_Previous_PE

Mapping to RIFT Configuration

The values defined in the YAML specification are processed by rift.py and the rift.ini template to produce the final execution environment.

Mapping Table

Actionable Configuration

When modifying the RIFT/asimov/rift.ini template, keep in mind that the low-level code (util_RIFT_pseudo_pipe.py) only consumes a subset of the provided keys.

Priority Actionable Items:

  • Approximant and Modes: Crucial for the underlying waveform generation.

  • Assumption Flags: Directly control the physics of the likelihood calculation.

  • CIP/ILE Parameters: Control the convergence and efficiency of the sampling process.

  • Priors: Ensure the search space is correctly bounded.

Everything else in the ini file is generally considered boilerplate or legacy filler.