Semi-Parametric Inducing Point Networks and Neural Processes

TL;DR

This paper presents SPIN, a semi-parametric neural architecture with linear complexity for efficient large-scale data querying, enhancing meta-learning and genotype imputation performance.

Contribution

Introduction of SPIN, a novel semi-parametric architecture with linear complexity, enabling efficient large-scale data querying for meta-learning and probabilistic modeling.

Findings

SPIN reduces memory requirements significantly.

SPIN improves accuracy across various meta-learning tasks.

SPIN achieves state-of-the-art results in genotype imputation.

Abstract

We introduce semi-parametric inducing point networks (SPIN), a general-purpose architecture that can query the training set at inference time in a compute-efficient manner. Semi-parametric architectures are typically more compact than parametric models, but their computational complexity is often quadratic. In contrast, SPIN attains linear complexity via a cross-attention mechanism between datapoints inspired by inducing point methods. Querying large training sets can be particularly useful in meta-learning, as it unlocks additional training signal, but often exceeds the scaling limits of existing models. We use SPIN as the basis of the Inducing Point Neural Process, a probabilistic model which supports large contexts in meta-learning and achieves high accuracy where existing models fail. In our experiments, SPIN reduces memory requirements, improves accuracy across a range of meta-learning tasks, and improves state-of-the-art performance on an important practical problem, genotype imputation.