# On the Number of Linear Regions of Deep Neural Networks

**Authors:** Guido Montúfar, Razvan Pascanu, Kyunghyun Cho, Yoshua Bengio  
**arXiv:** [1402.1869](https://arxiv.org/abs/1402.1869)  
**Date:** 2014-02-08  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1402.1869v1/

## Key points

- Paper title: "On the Number of Linear Regions of Deep Neural Networks"
- Authors: Guido Montúfar, Razvan Pascanu, Kyunghyun Cho, Yoshua Bengio
- Focus on complexity of functions computed by deep feedforward neural networks with piece-wise linear activations
- Analysis of number of regions of linearity within networks to understand impact of depth on performance
- Investigation of compositional maps to reveal advantages of depth in computational efficiency and capturing intricate patterns in data
- Contribution to understanding how deep neural networks use architecture for modeling complex functions through compositionality and computation reuse across layers

## Summary

The paper "On the Number of Linear Regions of Deep Neural Networks" by Guido Montúfar, Razvan Pascanu, Kyunghyun Cho, and Yoshua Bengio delves into the complexity of functions computed by deep feedforward neural networks with piece-wise linear activations. The study focuses on analyzing the number of regions of linearity within these networks to gain new theoretical insights into how depth impacts their performance. By investigating the compositional maps created by these networks, the authors shed light on the advantages that depth provides in terms of computational efficiency and capturing intricate patterns within data. This analysis contributes to a deeper understanding of how deep neural networks leverage their architecture to effectively model complex functions through compositionality and reuse of computations across layers.