Abstract:Graph partitioning aims to divide a graph into disjoint subsets while optimizing a specific partitioning objective. The majority of formulations related to graph partitioning exhibit NP-hardness due to their combinatorial nature. Conventional methods, like approximation algorithms or heuristics, are designed for distinct partitioning objectives and fail to achieve generalization across other important partitioning objectives. Recently machine learning-based methods have been developed that learn directly from data. Further, these methods have a distinct advantage of utilizing node features that carry additional information. However, these methods assume differentiability of target partitioning objective functions and cannot generalize for an unknown number of partitions, i.e., they assume the number of partitions is provided in advance. In this study, we develop NeuroCUT with two key innovations over previous methodologies. First, by leveraging a reinforcement learning-based framework over node representations derived from a graph neural network and positional features, NeuroCUT can accommodate any optimization objective, even those with non-differentiable functions. Second, we decouple the parameter space and the partition count making NeuroCUT inductive to any unseen number of partition, which is provided at query time. Through empirical evaluation, we demonstrate that NeuroCUT excels in identifying high-quality partitions, showcases strong generalization across a wide spectrum of partitioning objectives, and exhibits strong generalization to unseen partition count.

What problem does this paper attempt to address?

### Problems the paper attempts to solve The paper "NeuroCUT: A Neural Approach for Robust Graph Partitioning" aims to solve several key problems in graph partitioning: 1. **Non - inductiveness to the number of partitions**: - Existing neural methods need to specify the number of partitions during training and cannot generalize to unseen numbers of partitions. This means that if the number of partitions needs to be adjusted in practical applications, the model must be retrained. This is a significant limitation in practical applications such as chip design, because the optimal number of partitions is usually unknown in advance. 2. **Non - generalization of different objective functions**: - Graph partitioning problems can have different objective functions, such as Min Cut, Normalized Cut, Balanced Cut, and Sparsest Cut. Existing neural methods can usually only handle specific objective functions and cannot flexibly adapt to different objective functions. 3. **Differentiability assumptions of objective functions**: - Existing neural methods assume that the objective function is differentiable, but many objective functions in real - world problems (such as Sparsest Cut and Balanced Cut) are non - differentiable. This limits the applicability of these methods in real - world problems. ### Main contributions of the paper To solve the above problems, the paper proposes the NeuroCUT framework with the following main contributions: 1. **Diverse objective functions**: - NeuroCUT is an autoregressive graph reinforcement learning framework that integrates location information and can solve the graph partitioning problem of attributed graphs. Unlike traditional algorithms, NeuroCUT can handle multiple graph partitioning objective functions and does not require the differentiability of the objective function. 2. **Inductiveness to the number of partitions**: - The parameter space of NeuroCUT is independent of the number of partitions. This innovative decoupled architecture enables NeuroCUT to effectively generalize to any number of partitions. This allows the model to handle unseen numbers of partitions during inference without the need for retraining. 3. **Empirical evaluation**: - The authors conducted comprehensive experiments using real - world datasets to evaluate the performance of NeuroCUT on four different graph partitioning objective functions. Empirical studies show that NeuroCUT performs well in partitioning tasks and is robust to multiple objective functions. In addition, it also demonstrates the effective generalization ability of NeuroCUT to unseen numbers of partitions. ### Specific formulation of the problem Given a graph \(G\) and the number of partitions \(k\), the goal is to find a graph partition \(P\) such that an objective function \(O\_bj(G, P)\) is optimized. Specifically, NeuroCUT learns a policy \(\pi\) to assign each node \(v\in V\) to a partition. The policy \(\pi\) needs to satisfy the following properties: 1. **Inductiveness**: - The parameters of the policy \(\pi\) should be independent of the size of the graph and the number of partitions \(k\). If the policy is not inductive, it is impossible to infer unseen graph sizes or numbers of partitions. 2. **Learning diverse objective functions**: - To optimize the parameters of the policy \(\pi\), an objective function is required. The objective function may be non - differentiable, so the policy \(\pi\) should be able to learn to optimize non - differentiable objective functions. ### Summary NeuroCUT solves the limitations of existing methods in graph partitioning problems by introducing an autoregressive graph reinforcement learning framework, especially in terms of inductiveness to the number of partitions and generalization ability to different objective functions. This makes NeuroCUT more flexible and robust in practical applications.

NeuroCUT: A Neural Approach for Robust Graph Partitioning

Graph Partitioning and Sparse Matrix Ordering using Reinforcement Learning and Graph Neural Networks

GAP: Generalizable Approximate Graph Partitioning Framework

An end-to-end bi-objective approach to deep graph partitioning

Partition Clustering in Complex Weighted Networks Using K-Cut Ranking and Krill-Herd Optimization

CUTTANA: Scalable Graph Partitioning for Faster Distributed Graph Databases and Analytics

Deep Graph Reinforcement Learning for Solving Multicut Problem

Brief Announcement: Distributed Unconstrained Local Search for Multilevel Graph Partitioning

System partitioning on MCM using a new neural network model

An approximation algorithm for graph partitioning via deterministic annealing neural network

Deep Learning and Spectral Embedding for Graph Partitioning

Optimizing Graph Partition by Optimal Vertex-Cut: A Holistic Approach.

Erdos Goes Neural: an Unsupervised Learning Framework for Combinatorial Optimization on Graphs

Bipartite-Oriented Distributed Graph Partitioning for Big Learning

Multi-domain Network Slice Partitioning: A Graph Neural Network Algorithm

Uplifting the Expressive Power of Graph Neural Networks through Graph Partitioning

Distributed Constrained Combinatorial Optimization leveraging Hypergraph Neural Networks

Partitioner Selection with EASE to Optimize Distributed Graph Processing

VLSI Hypergraph Partitioning with Deep Learning

Highly Connected Graph Partitioning: Exact Formulation and Solution Methods