Linked Data Classification using Neurochaos Learning

Pooja Honna; Ayush Patravali; Nithin Nagaraj; Nanjangud C. Narendra

cs.LGFeb 18, 2026

Linked Data Classification using Neurochaos Learning

Authors: Pooja Honna, Ayush Patravali, Nithin Nagaraj, Nanjangud C. Narendra

Organizations: Complex Systems Programme National Institute of Advanced Studies (NIAS) IISc Campus, Bangalore 560 012, India

Abstract

Neurochaos Learning (NL) has shown promise in recent times over traditional deep learning due to its two key features: ability to learn from small sized training samples, and low compute requirements. In prior work, NL has been implemented and extensively tested on separable and time series data, and demonstrated its superior performance on both classification and regression tasks. In this paper, we investigate the next step in NL, viz., applying NL to linked data, in particular, data that is represented in the form of knowledge graphs. We integrate linked data into NL by implementing node aggregation on knowledge graphs, and then feeding the aggregated node features to the simplest NL architecture: ChaosNet. We demonstrate the results of our implementation on homophilic graph datasets as well as heterophilic graph datasets of verying heterophily. We show better efficacy of our approach on homophilic graphs than on heterophilic graphs. While doing so, we also present our analysis of the results, as well as suggestions for future work.

Figures & tables

Figure 1: Neurochaos-based feature transformation and classification pipeline.

Dataset	Nodes	Features	Edges	Classes	Hom.	Het.	Graph Property
Cora	2,708	1,433	5,429	7	0.810	0.190	Highly homophilic
Citeseer	3,327	3,703	4,732	6	0.738	0.262	Moderately homophilic
PubMed	19,717	500	44,338	3	0.802	0.198	Moderately homophilic
Actor	7,600	932	30,019	5	0.219	0.781	Highly heterophilic
Cornell	183	1,703	295	5	0.131	0.869	Highly heterophilic
Wisconsin	251	1,703	499	5	0.196	0.804	Highly heterophilic

Table 1: Quantitative summary of linked datasets used in this study.

Dataset	Graph Property	Q	B	$\epsilon$	Train Macro F1	Test Macro F1
Cora	Highly Homophilic	0.52	0.75	0.10	0.828	0.844
PubMed	Homophilic	0.524	0.759	0.0987	0.78	0.76
Citeseer	Homophilic	0.53	0.76	0.0985	0.705	0.732
Cornell	Partial Homo/Hetero	0.524	0.76	0.0985	0.53	0.43
Wisconsin	Partial Homo/Hetero	0.525	0.7625	0.0975	0.58	0.57
Actor	Highly Heterophilic	0.44	0.66	0.108	0.265	0.265

Table 2: Mean Node Aggregation Results on All Datasets

Dataset	Graph Property	Q	B	$\epsilon$	Train Macro F1	Test Macro F1
Cora	Homophilic	0.18	0.14	0.18	0.308	0.23
Actor	Heterophilic	0.44	0.66	0.108	0.232	0.25
Cornell	Partial Homo/Hetero	0.13	0.15	0.20	0.263	0.232

Table 3: H2GNN Results on Selected Datasets

Q	B	$\epsilon$	TrainMacro F1
0.52	0.75	0.10	0.8437
0.50	0.70	0.09	0.8130
0.49	0.75	0.07	0.8026

Table 4: Best ChaosNet Runs on the Cora Dataset

Figure 2: 3D visualization of macro F1-score across the ChaosNet hyperparameter space on the Cora dataset

Node Classification

Node classification assigns labels to graph entities by combining their features with information from connected nodes and graph structure.

See all 99 papers

GaLoRA: Parameter-Efficient Graph-Aware LLMs for Node Classification
Mar 11, 2026Mayur Choudhary, Saptarshi Sengupta, Katerina PotikaLLMs for Graph Learning Text-Attributed Graph Learning
Department of Computer Science, San Jose State University
Chimaera: A Mixture-of-Graph-Experts Architecture for Cross-Task and Cross-Dataset Graph Learning
Sep 8, 2026Jonatan Frank, David Richerby, Ansgar ScherpTransfer Learning Text-Attributed Graph Learning
University of Ulm, Ulm, Germany · University of Essex, Colchester, UK
Closed-Form Node Classification with Exact Graph Unlearning
May 25, 2026Aditya Gaur, Charu SharmaGraph Representation Learning Privacy-Preserving ML
Machine Learning Lab · IIIT Hyderabad

Mar 11, 2026J/K move · Enter open · S save

GaLoRA: Parameter-Efficient Graph-Aware LLMs for Node Classification

Mayur Choudhary, Saptarshi Sengupta, Katerina Potika

Department of Computer Science, San Jose State University

LLMs for Graph Learning Text-Attributed Graph Learning Node Classification Parameter-Efficient Fine-Tuning

Open paper AskPDF arXiv

The rapid rise of large language models (LLMs) and their ability to capture semantic relationships has led to their adoption in a wide range of applications. Text-attributed graphs (TAGs) are a notable example where LLMs can be combined with Graph Neural Networks to improve the performance of node classification. In TAGs, each node is associated with textual content and such graphs are commonly seen in various domains such as social networks, citation graphs, recommendation systems, etc. Effectively learning from TAGs would enable better representations of both structural and textual representations of the graph and improve decision-making in relevant domains. We present GaLoRA, a parameter-efficient framework that integrates structural information into LLMs. GaLoRA demonstrates competitive performance on node classification tasks with TAGs, performing on par with state-of-the-art models with just 0.24% of the parameter count required by full LLM fine-tuning. We experiment with three real-world datasets to showcase GaLoRA's effectiveness in combining structural and semantical information on TAGs.

Abstract

Node Classification

Explore similar work

GaLoRA: Parameter-Efficient Graph-Aware LLMs for Node Classification

Chimaera: A Mixture-of-Graph-Experts Architecture for Cross-Task and Cross-Dataset Graph Learning

Closed-Form Node Classification with Exact Graph Unlearning