• Corpus ID: 237396284

# ReLU Neural Networks of Polynomial Size for Exact Maximum Flow Computation

@inproceedings{Hertrich2021ReLUNN,
title={ReLU Neural Networks of Polynomial Size for Exact Maximum Flow Computation},
author={Christoph Hertrich and Leon Sering},
year={2021}
}
• Published 12 February 2021
• Computer Science
This paper studies the expressive power of artiﬁcial neural networks with rectiﬁed linear units. In order to study them as a model of real-valued computation, we introduce the concept of Max-Aﬃne Arithmetic Programs and show equivalence between them and neural networks concerning natural complexity measures. We then use this result to show that two fundamental combinatorial optimization problems can be solved with polynomial-size neural networks. First, we show that for any undirected graph…
1 Citations

## Figures from this paper

• Computer Science
ArXiv
• 2022
The algorithmic problem of finding the optimal weights and biases for a two-layer fully connected neural network to a given set of data points is considered and it is shown that even very simple networks are difficult to train.
• Computer Science
• 2022
The algorithmic problem of finding the optimal weights and biases for a two-layer fully connected neural network to a given set of data points is considered and it is shown that even very simple networks are difficult to train.

## References

SHOWING 1-10 OF 66 REFERENCES

• Computer Science
ArXiv
• 2022
The algorithmic problem of finding the optimal weights and biases for a two-layer fully connected neural network to a given set of data points is considered and it is shown that even very simple networks are difficult to train.
• Computer Science, Mathematics
IPCO
• 2022
This article precisely characterize the class of functions that are representable by such neural networks and shows that 2 hidden layers are necessary and suﬃcient to represent any function representable in the class, and proposes a new class of neural networks that is called shortcut linear threshold networks.
• Computer Science
Electron. Colloquium Comput. Complex.
• 2017
A study of high depth networks using ReLU gates which implement the function xmapsto x to understand the role of depth by showing size lowerbounds against such network architectures in parameter regimes hitherto unexplored.
• Computer Science
ITCS
• 2021
It is proved that under reasonable hardness assumptions, any proper learning algorithm for finding the best fitting ReLU must run in time exponential in $1/\epsilon^2$, which implies the first separation between proper and improper algorithms for learning a ReLU.
• Computer Science
J. Artif. Intell. Res.
• 2022
This work provides running time lower bounds in terms of W[1]-hardness for parameter d and proves that known brute-force strategies are essentially optimal (assuming the Exponential Time Hypothesis).
• Computer Science
ICML
• 2018
The results indicate that a deep rectifier network can only have more linear regions than every shallow counterpart with same number of neurons if that number exceeds the dimension of the input.
• Computer Science
NIPS
• 2014
We study the complexity of functions computable by deep feedforward neural networks with piecewise linear activations in terms of the symmetries and the number of linear regions that they have. Deep
• Computer Science
IEEE Global Telecommunications Conference GLOBECOM '91: Countdown to the New Millennium. Conference Record
• 1991
An attempt is made to show how to apply neural network optimization techniques to solve the resulting linear programming problem in real time, an extended version of the linear programming network proposed by L.O. Chua and G.N. Lin (1984, 1985).
• Computer Science, Mathematics
Found. Trends Theor. Comput. Sci.
• 2010
The goal of this monograph is to survey the field of arithmetic circuit complexity, focusing mainly on what it finds to be the most interesting and accessible research directions, with an emphasis on works from the last two decades.
• Computer Science, Mathematics
Electron. Colloquium Comput. Complex.
• 2017
The gap theorems hold for smoothly parametrized families of "hard" functions, contrary to countable, discrete families known in the literature, and a new lowerbound on the number of affine pieces is shown, larger than previous constructions in certain regimes of the network architecture.