{"success":true,"data":{"author":"Polymathic AI (The Well); Michael McCabe","bucketAccessType":"public","bucketName":"polymathic-ai/acoustic_scattering_maze","bucketPrefix":"","category":"Physics","cloudProvider":"huggingface","createdAt":"2026-11-14T01:01:19.253Z","description":"2D simulations of the variable-coefficient linear acoustic equations, which describe a pressure wave propagating through materials with different density and sound speed. Pressure waves leave point sources and travel through maze-like domains of low-density paths and walls orders of magnitude denser, a machine learning challenge with parallels to optimal placement problems such as WiFi coverage in a building. Solved with Clawpack on a uniform grid over [-1, 1] by [-1, 1], open in y with reflective walls in x; each trajectory has 201 steps of 256 by 256 images over t from 0 to 4. Fields are pressure, material density, material sound speed and velocity. The datasheet lists 2,000 trajectories. Listing data/ (measured 2026-11-13) shows train/ (16 files, 256 GB), valid/ and test/ (2 files, 32 GB each), stored as chunked files of many trajectories; about 320 GB in the repository. Files are HDF5 in the common The Well format; open them with the_well package (WellDataModule with base path hf://datasets/polymathic-ai/) or directly with h5py. Part of The Well (Ohana et al., NeurIPS 2024 Datasets and Benchmarks, arXiv:2412.00568); released under CC-BY-4.0, cite The Well and the simulation's source papers.","displayTitle":"The Well: Acoustic Scattering Through Mazes","id":"923d2f7f-e6cb-5e2c-9833-a7d6f6841dcf","license":"CC-BY-4.0","licenseEvidenceUrl":"https://arxiv.org/html/2412.00568","objectCountEstimate":31,"probeStatus":"ok","referenceUrl":"https://arxiv.org/abs/2412.00568","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"polymathic-ai/acoustic_scattering_maze\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"polymathic-ai/acoustic_scattering_maze\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=polymathic-ai, license:cc-by-4.0, downloads=1932"],"format":"mixed","generatedAt":"2026-11-14T01:01:20.437Z","layout":{"example":"data/test/acoustic_scattering_maze_chunk_18.hdf5","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/test/acoustic_scattering_maze_chunk_18.hdf5","truncatedAt":512}],"sampleObjects":["acoustic_scattering_maze.yaml","data/test/acoustic_scattering_maze_chunk_18.hdf5","data/test/acoustic_scattering_maze_chunk_19.hdf5","data/train/acoustic_scattering_maze_chunk_0.hdf5","data/train/acoustic_scattering_maze_chunk_1.hdf5"],"sidecars":[{"bytes":"---\nlanguage:\n- en\nlicense: cc-by-4.0\ntags:\n- physics\n\ntask_categories:\n- time-series-forecasting\n- other\ntask_ids:\n- multivariate-time-series-forecasting\n---\n\nThis Dataset is part of <a href=\"arxiv.org/abs/2412.00568\">The Well Collection</a>.\n\n# How To Load from HuggingFace Hub\n\n1. Be sure to have `the_well` installed (`pip install the_well`)\n2. Use the `WellDataModule` to retrieve data as follows:\n\n```python\nfrom the_well.benchmark.data import WellDataModule\n\n# The following line may take a couple of minutes to instantiate the datamodule\ndatamodule = WellDataModule(\n    \"hf://datasets/polymathic-ai/\",\n    \"acoustic_scattering_maze\",\n)\ntrain_dataloader = datamodule.train_dataloader()\n\nfor batch in dataloader:\n    # Process training batch\n    ...\n```\n\n# Acoustic Scattering - Maze\n\n**One line description of the data:** Simple acoustic wave propogation through maze-like structures.\n\n**Longer description of the data:** These variable-coefficient acoustic equations describe the propogation of an acoustic pressure wave through maze-like domains. Pressure waves emerge from point sources and propogate through domains consisting of low density maze paths and orders of magnitude higher density maze walls. This is built primarily as a challenge for machine learning methods, though has similar properties to optimal placement problems like WiFi in a building.\n\n**Domain expert**: [Michael McCabe](https://mikemccabe210.github.io/), Polymathic AI.\n\n**Code or software used to generate the data**: Clawpack,  adapted from [this example.](http://www.clawpack.org/gallery/pyclaw/gallery/acoustics_2d_interface.html)\n\n**Equation**:\n\n$$\n\\begin{align}\n\\frac{ \\partial p}{\\partial t} + K(x, y) \\left( \\frac{\\partial u}{\\partial x} + \\frac{\\partial v}{\\partial y} \\right) &= 0 \\\\\n\\frac{ \\partial u  }{\\partial t} + \\frac{1}{\\rho(x, y)} \\frac{\\partial p}{\\partial x} &= 0 \\\\\n\\frac{ \\partial v  }{\\partial t} + \\frac{1}{\\rho(x, y)} \\frac{\\partial p}{\\partial y} &= 0\n\\end{align}\n$$\n\nwith \\\\(\\rho\\\\) the material density, \\\\(u, v\\\\) the velocity in the \\\\(x, y\\\\) directions respectively, \\\\(p\\\\) the pressure, and \\\\(K\\\\) the bulk modulus.\n\nExample material densities can be seen below:\n\n![image](https://users.flatironinstitute.org/~polymathic/data/the_well/datasets/acoustic_scattering_maze/gif/mazes_density.png)\n\nTraversal can be seen:\n\n![Gif](https://users.flatironinstitute.org/~polymathic/data/the_well/datasets/acoustic_scattering_maze/gif/pressure_normalized.gif)\n\n| Dataset    | FNO | TFNO  | Unet | CNextU-net\n|:-:|:-:|:-:|:-:|:-:|\n| `acoustic_scattering_maze`  | 0.5062 | 0.5057| 0.0351| \\\\(\\mathbf{0.0153}\\\\)|\n\nTable: VRMSE metrics on test sets (lower is better). Best results are shown in bold. VRMSE is scaled such that predicting the mean value of the target field results in a score of 1.\n\n# About the data\n\n**Dimension of discretized data:** 201 steps of 256 \\\\(\\times\\\\) 256 images.\n\n**Fields available in the data:** pressure (scalar field), material density (constant scalar field), material speed of sound (constant scalar field), velocity field (vector field).\n\n**Number of trajectories:** 2000.\n\n**Estimated size of the ensemble of all simulations:** 311.3 GB.\n\n**Grid type:** uniform, cartesian coordinates.\n\n**Initial conditions:** Flat pressure static field with 1-6 high pressure rings randomly placed along paths of maze. The rings are defined with variable intensity \\\\(\\sim \\mathcal U(3., 5.)\\\\) and radius \\\\(\\sim \\mathcal U(.01, .04)\\\\). Any overlap with walls is removed.\n\n**Boundary conditions:** Open domain in \\\\(y\\\\), reflective walls in \\\\(x\\\\).\n\n**Simulation time-step:** Variable based on CFL with safety factor .25.\n\n**Data are stored separated by ( \\\\(\\Delta t\\\\)):** 2/201.\n\n**Total time range ( \\\\(t_{min}\\\\) to \\\\(t_{max}\\\\)):** [0, 4.].\n\n**Spatial domain size ( \\\\(L_x\\\\), \\\\(L_y\\\\)):** [-1, 1] x [-1, 1].\n\n**Set of coefficients or non-dimensional parameters evaluated:**\n\n- \\\\(K\\\\) is fixed at 4.0.\n\n- \\\\(\\rho\\\\) is the primary coefficient here. We generated a maze with initial width between 6 and 16 pixels and upsample it via nearest neighbor res","encoding":"utf8","key":"README.md","name":"README.md","size":5397,"truncated":true}],"sourceObjectKey":"data/test/acoustic_scattering_maze_chunk_18.hdf5","starterCell":"from huggingface_hub import hf_hub_download\npath = hf_hub_download(repo_id=\"polymathic-ai/acoustic_scattering_maze\", filename=\"data/test/acoustic_scattering_maze_chunk_18.hdf5\", repo_type=\"dataset\")\nprint(path)","starterSamplePath":"data/test/acoustic_scattering_maze_chunk_18.hdf5","totalObjects":31},"summary":"2D simulations of the variable-coefficient linear acoustic equations, which describe a pressure wave propagating through materials with different density and sound speed.","title":"The Well: Acoustic Scattering Through Mazes","updatedAt":"2026-11-14T02:40:10.930Z","visibility":"public","websiteUrl":"https://polymathic-ai.org/the_well/"}}