Methods

The methods and techniques used across SBI4GALEV, with how often each was mentioned in talk transcripts and slides, and the talks that covered it. Extracted and counted on-box.

Diffusion Models

11 transcript · 11 slide mentions · 7 talk(s)

Generative models that learn to reverse a gradual noising process to sample distributions.

Spectral Energy Distribution Fitting SED fitting

0 transcript · 19 slide mentions · 6 talk(s)

Process of modeling galaxy physical properties from their multi-wavelength flux.

Neural Posterior Estimation NPE

3 transcript · 10 slide mentions · 6 talk(s)

Amortized inference method that learns to map observations directly to the posterior distribution.

Hamiltonian Monte Carlo HMC

5 transcript · 5 slide mentions · 2 talk(s)

Gradient-based MCMC sampling method efficient for high-dimensional parameter spaces.

Maximum Mean Discrepancy MMD

3 transcript · 6 slide mentions · 2 talk(s)

A kernel-based distance metric used to compare two probability distributions.

Neural Likelihood Estimation NLE

3 transcript · 5 slide mentions · 3 talk(s)

Method that learns an emulator for the likelihood function to perform Bayesian inference.

Normalizing Flows

7 transcript · 0 slide mentions · 5 talk(s)

Generative models that transform simple distributions into complex ones via invertible mappings.

Nested Sampling

4 transcript · 3 slide mentions · 3 talk(s)

A Bayesian sampling method used for evidence calculation and multimodal posteriors.

Simulation-Based Calibration

3 transcript · 2 slide mentions · 4 talk(s)

Validation technique to check if inferred posteriors are statistically consistent with the truth.

Automatic Differentiation AD

1 transcript · 4 slide mentions · 2 talk(s)

Computational technique to automatically calculate derivatives of a function.

Neural Ratio Estimation NRE

1 transcript · 2 slide mentions · 2 talk(s)

Inference method that estimates the likelihood ratio to recover the posterior.

Likelihood-Free Inference LFI

1 transcript · 0 slide mentions · 1 talk(s)

Inference performed using simulators when the analytical likelihood is intractable.