# Towards a learning-based performance modeling for accelerating Deep Neural Networks

arXiv: https://arxiv.org/abs/2212.05031 · HTML: https://mlsanity.com/paper/2212.05031/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 22 references that are in the map.
- First posted on arXiv 9 Dec 2022.

## Builds on (most cited first, 8 of 8)

- [Rethinking the Inception Architecture for Computer Vision](https://mlsanity.com/paper/1512.00567/) (2015)
- [Parallel Multi Channel Convolution using General Matrix Multiplication](https://mlsanity.com/paper/1704.04428/) (2017)
- [CLTune: A Generic Auto-Tuner for OpenCL Kernels](https://mlsanity.com/paper/1703.06503/) (2017)
- [MEC: Memory-efficient Convolution for Deep Neural Network](https://mlsanity.com/paper/1706.06873/) (2017)
- [Machine Learning Based Auto-tuning for Enhanced OpenCL Performance Portability](https://mlsanity.com/paper/1506.00842/) (2015)
- [A model-driven approach for a new generation of adaptive libraries](https://mlsanity.com/paper/1806.07060/) (2018)
- [Algorithms and Heuristics for Scalable Betweenness Centrality Computation on Multi-GPU Systems](https://mlsanity.com/paper/1602.00963/) (2016)
- [Accelerating Energy Games Solvers on Modern Architectures](https://mlsanity.com/paper/1710.03647/) (2017)

## Abstract (arXiv)

Emerging applications such as Deep Learning are often data-driven, thus traditional approaches based on auto-tuners are not performance effective across the wide range of inputs used in practice. In the present paper, we start an investigation of predictive models based on machine learning techniques in order to optimize Convolution Neural Networks (CNNs). As a use-case, we focus on the ARM Compute Library which provides three different implementations of the convolution operator at different numeric precision. Starting from a collation of benchmarks, we build and validate models learned by Decision Tree and naive Bayesian classifier. Preliminary experiments on Midgard-based ARM Mali GPU show that our predictive model outperforms all the convolution operators manually selected by the library.