Conditional Density Estimation Tools in Python and R with Applications to Photometric Redshifts and Likelihood-Free Cosmological Inference

Abstract

It is well known in astronomy that propagating non-Gaussian prediction uncertainty in photometric redshift estimates is key to reducing bias in downstream cosmological analyses. Similarly, likelihood-free inference approaches, which are beginning to emerge as a tool for cosmological analysis, require a characterization of the full uncertainty landscape of the parameters of interest given observed data. However, most machine learning (ML) or training-based methods with open-source software target point prediction or classification, and hence fall short in quantifying uncertainty in complex regression and parameter inference settings. As an alternative to methods that focus on predicting the response (or parameters) $\mathbf{y}$ from features $\mathbf{x}$, we provide nonparametric conditional density estimation (CDE) tools for approximating and validating the entire probability density function (PDF) $\mathrm{p}(\mathbf{y}|\mathbf{x})$ of $\mathbf{y}$ given (i.e., conditional on) $\mathbf{x}$. As there is no one-size-fits-all CDE method, the goal of this work is to provide a comprehensive range of statistical tools and open-source software for nonparametric CDE and method assessment which can accommodate different types of settings and be easily fit to the problem at hand. Specifically, we introduce four CDE software packages in $\text{Python}$ and $\text{R}$ based on ML prediction methods adapted and optimized for CDE: $\text{NNKCDE}$, $\text{RFCDE}$, $\text{FlexCode}$, and $\text{DeepCDE}$. Furthermore, we present the $\text{cdetools}$ package, which includes functions for computing a CDE loss function for tuning and assessing the quality of individual PDFs, along with diagnostic functions. We provide sample code in $\text{Python}$ and $\text{R}$ as well as examples of applications to photometric redshift estimation and likelihood-free cosmological inference via CDE.

27 pages, 7 figures, 4 tables

References in corpus (44)

• Adam: A Method for Stochastic Optimization
• Scikit-learn: Machine Learning in Python
• XGBoost: A Scalable Tree Boosting System
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• LSST: from Science Drivers to Reference Design and Anticipated Data Products
• The Eleventh and Twelfth Data Releases of the Sloan Digital Sky Survey: Final Data from SDSS-III
• Weak Gravitational Lensing and its Cosmological Applications
• Weak lensing for precision cosmology
• Data Mining and Machine Learning in Astronomy
• A Critical Assessment of Photometric Redshift Methods: A CANDELS Investigation
• Cosmology with Weak Lensing Surveys
• KiDS-450 + 2dFLenS: Cosmological parameter constraints from weak gravitational lensing tomography and overlapping redshift-space galaxy clustering
• Introduction to astroML: Machine Learning for Astrophysics
• TPZ : Photometric redshift PDFs and ancillary information by using prediction trees and random forests
• KiDS+GAMA: Cosmology constraints from a joint analysis of cosmic shear, galaxy-galaxy lensing and angular clustering
• Fast likelihood-free cosmology with neural density estimators and active learning
• Photometric redshifts from SDSS images using a Convolutional Neural Network
• Photometric redshift estimation via deep learning
• Massive optimal data compression and density estimation for scalable, likelihood-free inference in cosmology
• Estimating the Redshift Distribution of Photometric Galaxy Samples II. Applications and Tests of a New Method
• Precision photometric redshift calibration for galaxy-galaxy weak lensing
• Fast $\epsilon$-free Inference of Simulation Models with Bayesian Conditional Density Estimation
• Dark Energy Survey Year 1 Results: Multi-Probe Methodology and Simulated Likelihood Analyses
• Cosmological Constraints from Multiple Probes in the Dark Energy Survey
• SKYNET: an efficient and robust neural network training tool for machine learning in astronomy
• Flexible statistical inference for mechanistic models of neural dynamics
• Evaluation of probabilistic photometric redshift estimation approaches for The Rubin Observatory Legacy Survey of Space and Time (LSST)
• Likelihood-free inference with emulator networks
• Sequential Neural Likelihood: Fast Likelihood-free Inference with Autoregressive Flows
• On the realistic validation of photometric redshifts, or why Teddy will never be Happy
• Nonparametric Conditional Density Estimation in a High-Dimensional Regression Setting
• Photometric Redshift Probability Distributions for Galaxies in the SDSS DR8
• What Lies Beneath: Using p(z) to Reduce Systematic Photometric Redshift Errors
• Photometric Redshift Estimation Using Spectral Connectivity Analysis
• A Spectral Series Approach to High-Dimensional Nonparametric Regression
• The Role of Machine Learning in the Next Decade of Cosmology
• ABC-CDE: Towards Approximate Bayesian Computation with Complex High-Dimensional Data and Limited Simulations
• Approximating photo-$z$ PDFs for large surveys
• A Unified Framework for Constructing, Tuning and Assessing Photometric Redshift Density Estimates in a Selection Bias Setting
• Sparse Representation of Photometric Redshift PDFs: Preparing for Petascale Astronomy
• High-Dimensional Density Ratio Estimation with Extensions to Approximate Likelihood Computation
• RFCDE: Random Forests for Conditional Density Estimation
• Uncertain Photometric Redshifts
• (f)RFCDE: Random Forests for Conditional Density Estimation and Functional Data

Cited by in corpus (14)

• Photometric Redshifts for Next-Generation Surveys
• Evaluation of probabilistic photometric redshift estimation approaches for The Rubin Observatory Legacy Survey of Space and Time (LSST)
• Weak Lensing Tomographic Redshift Distribution Inference for the Hyper Suprime-Cam Subaru Strategic Program three-year shape catalogue
• Photometric Redshifts from SDSS Images with an Interpretable Deep Capsule Network
• Spotting Anomalous Trades in NFT Markets: The Case of NBA Topshot
• Flexible distribution-free conditional predictive bands using density estimators
• A Composite Likelihood Approach for Inference under Photometric Redshift Uncertainty
• CD-split and HPD-split: efficient conformal regions in high dimensions
• Improving Photometric Redshift Estimates with Training Sample Augmentation
• The Quasar Catalogue for S-PLUS DR4 (QuCatS) and the estimation of photometric redshifts
• A Deep Generative Approach to Conditional Sampling
• Quantifying Epistemic Uncertainty in Deep Learning
• Re-calibrating Photometric Redshift Probability Distributions Using Feature-space Regression
• Photometric Redshift Estimation Using Scaled Ensemble Learning