# Measuring the Data Efficiency of Deep Learning Methods

arXiv: https://arxiv.org/abs/1907.02549 · HTML: https://mlsanity.com/paper/1907.02549/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 3 Jul 2019.

## Built on by (most cited first, 5 of 5)

- [Curriculum Learning for Vision-and-Language Navigation](https://mlsanity.com/paper/2111.07228/) (2021)
- [Reward prediction for representation learning and reward shaping](https://mlsanity.com/paper/2105.03172/) (2021)
- [Evaluation of Seismic Artificial Intelligence with Uncertainty](https://mlsanity.com/paper/2501.14809/) (2025)
- [Towards Sample-efficient Apprenticeship Learning from Suboptimal Demonstration](https://mlsanity.com/paper/2110.04347/) (2021)
- [Semi-self-supervised Automated ICD Coding](https://mlsanity.com/paper/2205.10088/) (2022)

## Abstract (arXiv)

In this paper, we propose a new experimental protocol and use it to benchmark the data efficiency --- performance as a function of training set size --- of two deep learning algorithms, convolutional neural networks (CNNs) and hierarchical information-preserving graph-based slow feature analysis (HiGSFA), for tasks in classification and transfer learning scenarios. The algorithms are trained on different-sized subsets of the MNIST and Omniglot data sets. HiGSFA outperforms standard CNN networks when the models are trained on 50 and 200 samples per class for MNIST classification. In other cases, the CNNs perform better. The results suggest that there are cases where greedy, locally optimal bottom-up learning is equally or more powerful than global gradient-based learning.