An open dataset of neural networks for hypernetwork research

2025-07-15

AI Summary

Current hypernetwork research is hindered by the absence of publicly available, structured neural network datasets. To address this, we introduce the first open-source dataset specifically designed for hypernetwork research: a collection of 10⁴ LeNet-5 models trained on a binary classification task over two classes from ImageNette V2 (1,000 distinct, verifiably separable models per class). All models were trained uniformly on a distributed cluster (>10⁴ CPU cores), ensuring consistent weight distribution and label verifiability. This work establishes the first large-scale, fine-grained, instance-level annotation of neural networks, enabling supervised model identification and hypernetwork-based weight generation. On the neural architecture classification task, our dataset achieves 72.0% accuracy—demonstrating that learnable structural distinctions exist across neural network instances. The complete dataset, generation scripts, and training configurations are publicly released.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Large Vision Models
• Natural Language Processing: (Large) Language Models

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Despite the transformative potential of AI, the concept of neural networks that can produce other neural networks by generating model weights (hypernetworks) has been largely understudied. One of the possible reasons is the lack of available research resources that can be used for the purpose of hypernetwork research. Here we describe a dataset of neural networks, designed for the purpose of hypernetworks research. The dataset includes $10^4$ LeNet-5 neural networks trained for binary image classification separated into 10 classes, such that each class contains 1,000 different neural networks that can identify a certain ImageNette V2 class from all other classes. A computing cluster of over $10^4$ cores was used to generate the dataset. Basic classification results show that the neural networks can be classified with accuracy of 72.0%, indicating that the differences between the neural networks can be identified by supervised machine learning algorithms. The ultimate purpose of the dataset is to enable hypernetworks research. The dataset and the code that generates it are open and accessible to the public.

Problem

• Lack of open datasets for hypernetwork research
• Need for neural networks generating other neural networks
• Limited resources to study weight-generated hypernetworks

Innovation

• Open dataset of 10^4 LeNet-5 neural networks
• Cluster with 10^4 cores generated dataset
• 72% accuracy in classifying neural networks