# On Deep Set Learning and the Choice of Aggregations

arXiv: https://arxiv.org/abs/1903.07348 · HTML: https://mlsanity.com/paper/1903.07348/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 18 Mar 2019.

## Built on by (most cited first, 13 of 13)

- [Likelihood-Free Parameter Estimation with Neural Bayes Estimators](https://mlsanity.com/paper/2208.12942/) (2022)
- [The Neural Process Family: Survey, Applications and Perspectives](https://mlsanity.com/paper/2209.00517/) (2022)
- [On permutation-invariant neural networks](https://mlsanity.com/paper/2403.17410/) (2024)
- [Postprocessing of Ensemble Weather Forecasts Using Permutation-invariant Neural Networks](https://mlsanity.com/paper/2309.04452/) (2023)
- [Robust Point Cloud Processing through Positional Embedding](https://mlsanity.com/paper/2309.00339/) (2023)
- [Design Strategy Network: A deep hierarchical framework to represent generative design strategies in complex action spaces](https://mlsanity.com/paper/2110.03760/) (2021)
- [Optimal sequential decision making with probabilistic digital twins](https://mlsanity.com/paper/2103.07405/) (2021)
- [SHIFT15M: Fashion-specific dataset for set-to-set matching with several distribution shifts](https://mlsanity.com/paper/2108.12992/) (2021)
- [Effective and Interpretable Information Aggregation with Capacity Networks](https://mlsanity.com/paper/2207.12013/) (2022)
- [Uniform Ck Approximation of G-Invariant and Antisymmetric Functions, Embedding Dimensions, and Polynomial Representations](https://mlsanity.com/paper/2403.01339/) (2024)
- [Improving Set Function Approximation with Quasi-Arithmetic Neural Networks](https://mlsanity.com/paper/2602.04941/) (2026)
- [MaxSketch: Robust Distinct Counting in Streams via Random Projections](https://mlsanity.com/paper/2605.15571/) (2026)
- [Population-Aware Physics-Informed Neural Particle Flow for Robust Spacecraft Bayesian Navigation](https://mlsanity.com/paper/2606.10959/) (2026)

## Abstract (arXiv)

Recently, it has been shown that many functions on sets can be represented by sum decompositions. These decompositons easily lend themselves to neural approximations, extending the applicability of neural nets to set-valued inputs---Deep Set learning. This work investigates a core component of Deep Set architecture: aggregation functions. We suggest and examine alternatives to commonly used aggregation functions, including learnable recurrent aggregation functions. Empirically, we show that the Deep Set networks are highly sensitive to the choice of aggregation functions: beyond improved performance, we find that learnable aggregations lower hyper-parameter sensitivity and generalize better to out-of-distribution input size.