Irina Higgins, Shakir Mohamed, constrained International 2017.

In contrast, the mathematics needed to analyze such schemes is what forms the focus in Stochastic Approximation (SA) theory [2, 4]. More generally, SA refers to an iterative scheme that helps find zeroes or optimal points of a function, for which only noisy evaluationsarepossible.

However, the simple power-law relation becomes more complicated when compute is considered.

Inthisparadigm, wepermit thesystem to defer a subset of its tasks to the expert. Although there are currently systems that follow this paradigm and are designed to optimize the accuracy of the final human-AI team, the general methodology for developing such systems under a set of constraints (e.g., algorithmic fairness, expert intervention budget, defer of anomaly,etc.)

K isthenumber ofpatches that are grouped from the input points.

Existing works usually estimate the missing shape by decoding a latent feature encoded from the input points.