GaMPEN: A Machine-learning Framework for Estimating Bayesian Posteriors of Galaxy Morphological Parameters

We introduce a novel machine learning framework for estimating the Bayesian posteriors of morphological parameters arbitrarily large numbers galaxies. The Galaxy Morphology Posterior Estimation Network (GaMPEN) estimates values and uncertainties galaxy's bulge-to-total light ratio ($L_B/L_T$), effective radius ($R_e$), flux ($F$). To estimate posteriors, GaMPEN uses Monte Carlo Dropout technique incorporates full covariance matrix between output in its loss function. also Spatial Transformer (STN) to automatically crop input galaxy frames an optimal size before determining their morphology. This will allow it be applied new data without prior knowledge size. Training testing on galaxies simulated match $z < 0.25$ Hyper Suprime-Cam Wide $g$-band images, we demonstrate that achieves typical errors $0.1$ $L_B/L_T$, $0.17$ arcsec ($\sim 7\%$) $R_e$, $6.3\times10^4$ nJy 1\%$) $F$. GaMPEN's predicted are well-calibrated accurate ($<5\%$ deviation) -- regions parameter space with high residuals, correctly predicts correspondingly uncertainties. can apply categorical labels (i.e., classifications such as "highly bulge-dominated") predictions residuals verify those $\gtrsim 97\%$ accurate. best our knowledge, is first joint posterior distributions multiple application STN optical imaging astronomy.