Graph Learning

Friendster is an on-line gaming network. Before re-launching as a game website, Friendster was a social networking site where users can form friendship edge each other. Friendster social network also allows users form a group which other members can then join. The Friendster dataset consist of ground-truth communities (based on user-defined groups) and the social network from induced subgraph of the nodes that either belong to at least one community or are connected to other nodes that belong to at least one community. Source: https://snap.stanford.edu/data/com-Friendster.html

A social network of LastFM users which was collected from the public API in March 2020. Nodes are LastFM users from Asian countries and edges are mutual follower relationships between them. The vertex features are extracted based on the artists liked by the users. The task related to the graph is multinomial node classification - one has to predict the location of users. This target feature was derived from the country field for each user.

The BotNet dataset is a set of topological botnet detection datasets forgraph neural networks. Source: https://github.com/harvardnlp/botnet-detection Image Source: https://github.com/harvardnlp/botnet-detection

Mindboggle is a large publicly available dataset of manually labeled brain MRI. It consists of 101 subjects collected from different sites, with cortical meshes varying from 102K to 185K vertices. Each brain surface contains 25 or 31 manually labeled parcels. Source: Graph Convolutions on Spectral Embeddings: Learning of Cortical Surface Data Image Source: https://mindboggle.info/data.html

Question Answering (QA) is a widely-used framework for developing and evaluating an intelligent machine. In this light, QA on Electronic Health Records (EHR), namely EHR QA, can work as a crucial milestone toward developing an intelligent agent in healthcare. EHR data are typically stored in a relational database, which can also be converted to a directed acyclic graph, allowing two approaches for EHR QA: Table-based QA and Knowledge Graph-based QA. MIMIC-SPARQL dataset provides graph-based EHR QA data where natural language queries are converted to SPARQL instead of SQL

ChEMBL is a manually curated database of bioactive molecules with drug-like properties. It brings together chemical, bioactivity and genomic data to aid the translation of genomic information into effective new drugs.

GVLQA is the first vision-language QA dataset for general graph reasoning. Contains a base set GVLQA-BASE and four image-augmented subsets GVLQA-AUGLY, GVLQA-AUGNO, GVLQA-AUGNS, GVLQA-AUGET, where the samples are relatively corresponding with the base set. Contains 7 graph reasoning tasks: detecting cycle, connectivity, computing topological ordering, shortest path, maximum flow, bipartite matching num, and Hamilton path. Utility: 1) evaluate the graph reasoning capabilities of VLMs or LLMs; 2) help models acquire fundamental graph comprehension and reasoning abilities as a pretraining dataset.

AsEP is a protein structure dataset that includes 1723 filtered antibody-antigen complexes from abYbank/AbDb. It is designed to facilitate the development of machine-learning-based models for predicting antibody-specific epitopes given a pair of antibody and antigen structures as input, which is important for antibody engineering. The dataset also introduces a new task for predicting the interacting residue pairs between antibodies and antigens, which is useful for understanding the functional properties of antibodies and enhances the interpretability of model performance. Please refer to this GitHub repository AsEP-dataset for the dataset interface, which is built with Python and PyTorch-Geometric Dataset Module.

The ETHZ Shape dataset contains images of five diverse shape-based classes, collected from Flickr and Google Images. The main challenges it offers are clutter, intra-class shape variability, and scale changes. The authors deliberately selected several images where the object comprises only a rather small portion of the image, and made an effort to include objects appearing at a wide range of scales. The objects are mostly unoccluded and are all taken from approximately the same viewpoint (the side). Source: http://calvin-vision.net/datasets/ethz-shape-classes/ Image Source: http://calvin-vision.net/datasets/ethz-shape-classes/

This dataset contains simulated synthetic particle decays, simulated using the PhaseSpace library. All simulated decay topologies have a common root particle of mass 100 (arbitrary units). Intermediate particles are selected at random with replacement from the following masses: [90, 80, 70, 50, 25, 20, 10]. Final state particles, which make up the leaf nodes of generated topologies, are drawn with replacement from the following masses: [1, 2, 3, 5, 12]. For each intermediate particle (including the root), we limit the minimum number of children to two, and the maximum five. The dataset contains the resulting simulated particle physics decays, with information about the detected particle (leaves) to be used as input, and Lowest Common Ancestor Generations (LCAGs) to be used as training targets. Tree topology creation to generate the dataset was as follows: starting from the root particle a set of children are selected from the available intermediate and final state particles such that the sum of their masses totals less than the root, this process is then repeated for each child particle which is not a final state particle and so on until only final state particles remain. This dataset consists of 200 topologies (unique decay processes) in total, with 16,000 samples per topology. In the paper's experiments, 2000 topologies for each of training, validation, and testing were used. Leaf node features are not normalized. We have not enforced any ordering of the nodes and leave them unsorted as created in the dataset.

The dataset covers the 2022-23 NBA regular season (2022-10-18 to 2023-01-20) which contains 691 games in 92 game days. There are 582 active players among the 30 teams. Besides 7 basic statistics, we collected 3 tracking statistics, and 3 advanced statistics. We use tracking statistics to more accurately reflect players' movements on the court, and advanced statistics to more properly represent a player's effectiveness and contribution to the game. Together, these two types of data give us a better understanding of factors that are not visible on the scoreboard.

OCB contains two graph datasets, Ckt-Bench-101 and Ckt-Bench-301, for representation learning over analog circuits. Ckt-Bench-101 and Ckt-Bench-301 contain graphs (DAGs) that represent analog circuits and provide their corresponding graph-level properties: DC gain (Gain), bandwidth (BW), phase margin (PM),Figure of Merit (FoM), which characterize the circuit performance. * Motivation: Facilitate research in representation learning and automation of analog circuits. * Tasks: graph-level prediction/regression; analog circuit search (ACS). * Node features: discrete feature that characterize device type (e.g. capacitor); and continues feature of the device (e.g. capacitance). * First open source benchmark for graph learning in analog circuits.

The analysis of building models for usable area, building safety, and energy efficiency requires accurate classification data of spaces and space elements. To reduce input model preparation effort and errors, automated classification of spaces and space elements is desirable. Although existing space function classifiers use space adjacency or connectivity graphs as input, the application of Graph Deep Learning (GDL) to space layout element classification has not been extensively researched due to the lack of suitable datasets. To bridge this gap, we introduce a dataset named SAGC-A68, which comprises access graphs automatically generated from 68 digital 3D models of space layouts of apartment buildings designed or built between 1952 and 2019 in 13 countries. Each access graph contains nodes representing spaces and space elements and edges representing the connection between them. Nodes are uniquely identified and characterized by 16 features including “Position X”, “Position Y”, “Position Z”, “Width”, “Height”, “Depth”, “Area”, “Volume”, “Is_internal”, “Door_opening_quantity”, “Window_quantity”, “Max_door_width”,” Encloses_ws”, “Is_contained_in_ws”, ”bounding_box”, and “Label” (28 identified labels are shown in bold type in Table 1). Edges are identified by a unique ID and characterized by three features, including “Z_angle”, “Delta_z”, and “Length”. In total, the dataset comprises 4871 nodes and 4566 edges, including disconnected nodes representing shafts. It is suitable for developing GDL models for space element and space function classification in Building information modeling (BIM) authoring systems.