Position: Topological Deep Learning is the New Frontier for Relational Learning
Authors: Theodore Papamarkou, Tolga Birdal, Michael M. Bronstein, Gunnar E. Carlsson, Justin Curry, Yue Gao, Mustafa Hajij, Roland Kwitt, Pietro Lio, Paolo Di Lorenzo, Vasileios Maroulas, Nina Miolane, Farzana Nasrin, Karthikeyan Natesan Ramamurthy, Bastian Rieck, Simone Scardapane, Michael T Schaub, Petar Veličković, Bei Wang, Yusu Wang, Guowei Wei, Ghada Zamzmi
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | Topological deep learning (TDL) is a rapidly evolving field that uses topological features to understand and design deep learning models. This paper posits that TDL is the new frontier for relational learning. TDL may complement graph representation learning and geometric deep learning by incorporating topological concepts, and can thus provide a natural choice for various machine learning settings. To this end, this paper discusses open problems in TDL, ranging from practical benefits to theoretical foundations. For each problem, it outlines potential solutions and future research opportunities. At the same time, this paper serves as an invitation to the scientific community to actively participate in TDL research to unlock the potential of this emerging field. |
| Researcher Affiliation | Collaboration | Theodore Papamarkou 1 Tolga Birdal 2 Michael Bronstein 3 Gunnar Carlsson 4 5 Justin Curry 6 Yue Gao 7 Mustafa Hajij 8 Roland Kwitt 9 Pietro Li o 10 Paolo Di Lorenzo 11 Vasileios Maroulas 12 Nina Miolane 13 Farzana Nasrin 14 Karthikeyan Natesan Ramamurthy 15 Bastian Rieck 16 17 Simone Scardapane 11 Michael T. Schaub 18 Petar Veliˇckovi c 19 10 Bei Wang 20 Yusu Wang 21 Guo-Wei Wei 22 Ghada Zamzmi 23 1Department of Mathematics, The University of Manchester, Manchester, UK. 2Department of Computing, Imperial College London, London, UK. 3Department of Computer Science, University of Oxford, Oxford, UK. 4Department of Mathematics, Stanford University, Stanford, USA. 5Blue Light AI Inc, USA. 6University at Albany, New York, USA. 7School of Software, Tsinghua University, Beijing, China. 8University of San Francisco, San Francisco, USA. 9Department of Artificial Intelligence and Human Interfaces, University of Salzburg, Austria. 10Department of Computer Science and Technology, University of Cambridge, Cambridge, UK. 11Department of Information Engineering, Electronics and Telecommunications, Sapienza University of Rome, Rome, Italy. 12Department of Mathematics, University of Tennessee, Knoxville, USA. 13Department of Electrical and Computer Engineering, UC Santa Barbara, Santa Barbara, USA. 14Department of Mathematics, University of Hawai i at M anoa, Hawai i, USA. 15IBM Corporation New York, USA. 16Helmholtz Munich, Munich Germany. 17Technical University of Munich, Munich Germany. 18RWTH Aachen University, Aachen, Germany. 19Google Deep Mind. 20School of Computing, University of Utah, Utah, USA. 21Computer Science and Engineering Department, University of California San Diego, San Diego, USA. 22Department of Mathematics, Michigan State University, East Lansing, Michigan, USA. 23University of South Florida, Florida, USA. |
| Pseudocode | No | The paper is a position paper discussing a field and open problems; it does not present new algorithms or methods that would require pseudocode or algorithm blocks. |
| Open Source Code | No | The paper discusses existing software packages for TDL (e.g., 'Topo X is a suite of Python packages designed to compute and learn with topological neural networks'), but it does not provide source code for any methodology or findings presented in *this specific paper*, as it is a survey/position paper. |
| Open Datasets | No | The paper discusses the need for datasets in TDL research, mentioning 'Open Graph Benchmark' as an example of existing benchmarks for graph machine learning, and states 'There is a scarcity of higher-order data'. However, it does not use specific datasets for its own experimental evaluation, as it is a position paper, and therefore does not provide concrete access information for a dataset used in its own research. |
| Dataset Splits | No | The paper is a position paper and does not describe experimental results or dataset splits used by its authors for training, validation, or testing. |
| Hardware Specification | No | The paper is a position paper and does not describe experiments conducted by its authors, thus no hardware specifications for running experiments are provided. |
| Software Dependencies | No | The paper discusses various software packages relevant to TDL (e.g., 'Network X (Hagberg et al., 2008)', 'Topo X (Hajij et al., 2024)'), but these are external tools or projects. The paper itself does not report on experiments requiring specific software dependencies for its own work, as it is a position paper. |
| Experiment Setup | No | The paper is a position paper and does not describe experiments conducted by its authors, thus no experimental setup details like hyperparameters or training settings are provided. |