{"success":true,"data":{"author":"Matthew Golden","bucketAccessType":"public","bucketName":"mgolden30/Kolmogorov_TW","bucketPrefix":"","category":"Physics","cloudProvider":"huggingface","createdAt":"2026-11-21T20:46:01.249Z","description":"A database of exact steady and relative-equilibrium solutions of 2D Kolmogorov flow, the incompressible Navier-Stokes equations driven by a sinusoidal body force, at the parameters common in the literature: viscosity nu = 1/40, vorticity forcing -4cos(4y), periodic domain [0, 2pi]^2, 128 x 128 resolution. The single file solutions.h5 holds 8,952 solutions, 847 equilibria (EQ) and 8,105 traveling waves (TW, relative equilibria) with their wave speeds. Solutions are grouped into 416 shape families (named F1 to F416 by ascending mean number of unstable directions) by recursive HDBSCAN clustering on a symmetry-reduced L2 distance that minimizes over the 16 discrete symmetries and continuous x-translation; 658 unclustered solutions are labeled N and family -1. Members of a family are aligned to the family reference solution. Read with h5py; helper Python scripts are included. Use it for studies of exact coherent structures, unstable periodic and invariant solutions underlying turbulence, and dynamical-systems approaches to chaotic fluid flow. Released under CC BY 4.0; cite the associated paper \"A catalog of exact steady states in Kolmogorov flow\" (arXiv:2609.27062).","displayTitle":"Kolmogorov Flow Exact Coherent States: 8,952 Equilibria and Traveling Waves of 2D Navier-Stokes (nu = 1/40)","huggingFaceUrl":"https://huggingface.co/datasets/mgolden30/Kolmogorov_TW","id":"7bda65c9-75e5-50fa-8aa9-bed7964502e0","license":"CC-BY-4.0","licenseEvidenceUrl":"https://huggingface.co/datasets/mgolden30/Kolmogorov_TW","objectCountEstimate":6,"probeStatus":"ok","referenceUrl":"https://arxiv.org/abs/2609.27062","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"mgolden30/Kolmogorov_TW\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"mgolden30/Kolmogorov_TW\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=mgolden30, license:cc-by-4.0, downloads=160"],"format":"mixed","generatedAt":"2026-11-21T20:48:52.248Z","layout":{"example":"solutions.h5","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"solutions.h5","truncatedAt":512}],"sampleObjects":["example.py","solutions.h5","tw_objective.py"],"sidecars":[{"bytes":"This work is licensed under the Creative Commons Attribution 4.0\nInternational License (CC BY 4.0).\n\nTo view a copy of this license, visit:\nhttps://creativecommons.org/licenses/by/4.0/\n\nYou are free to:\n  - Share -- copy and redistribute the material in any medium or format\n  - Adapt -- remix, transform, and build upon the material\n  for any purpose, even commercially.\n\nUnder the following terms:\n  - Attribution -- You must give appropriate credit, provide a link to\n    the license, and indicate if changes were made.\n\nNo additional restrictions -- you may not apply legal terms or\ntechnological measures that legally restrict others from doing anything\nthe license permits.\n","encoding":"utf8","key":"LICENSE","name":"LICENSE","size":680,"truncated":false},{"bytes":"---\nlicense: cc-by-4.0\n---\n\n## Kolmogorov Flow Traveling Wave / Equilibria Database\n\nEquilibria (EQ) and traveling waves (TW, also called relative equilibria) for\n2D Kolmogorov flow at the parameters commonly used in the literature: nu=1/40,\nvorticity forcing -4cos(4y), periodic domain [0,2pi]^2, 128x128 resolution.\n\n**8,952 solutions** (847 equilibria, 8,105 traveling waves), in a single file,\n`solutions.h5`.\n\n### Naming\n\nSolutions are named by shape family, not by discovery order: `F<family>-<n>`,\ne.g. `F17-3` is the 3rd solution in family 17. Families come from a\nsymmetry-invariant shape clustering: recursive application of HDBSCAN\n(Campello, Moulavi & Sander 2013; `sklearn.cluster.HDBSCAN`,\n`metric=\"precomputed\"`, `cluster_selection_method=\"leaf\"`, `min_cluster_size=10`)\nover the full pairwise distance matrix, where distance between two solutions\nis their field L2 distance minimized over the problem's 16 discrete\nsymmetries and continuous x-translation (a genuine metric -- every symmetry\noperation is an isometry, so this distance satisfies the triangle\ninequality). A single HDBSCAN pass leaves many solutions unclustered; we\nrepeatedly recluster whatever's left over until a pass finds nothing\nfurther, which took 39 levels and produced **416 families**, ranked\n**F1..F416 by ascending mean number of unstable directions** (F1 is the\nmost dynamically stable family).\n\n**`N-<n>` solutions (658 of 8952, 7.3%) are genuine HDBSCAN noise** -- the\nrecursive clustering never placed them in any cluster, at any of the 39\nlevels. This is expected, normal HDBSCAN behavior (it is a density-based\nmethod that is supposed to leave outliers as noise, unlike e.g. k-means\nwhich forces every point into some cluster) and is not folded into any\nfamily: `family == -1` for these, distinctly separate from `family` values\n`1..416`. **We do not claim `N` solutions belong anywhere** -- do not treat\nthem as a 417th \"family,\" and do not assume two `N` solutions are similar\nto each other just because they share the `N` prefix; they are simply\n\"not classified,\" each for its own reason.\n\n(Historical note, for anyone cross-referencing a build from before\n2026-09-19: an earlier version of this file folded those 658 noise\nsolutions into the family of their nearest already-classified neighbor --\npurely so every solution had a browsable name -- and labeled the resulting\ncatch-all group `U`. That group's reported size, 725, was never a genuine\nHDBSCAN cluster size: it was 628 real cluster members (now family `F253`)\nplus 97 of the noise solutions glued on afterward for naming convenience.\nThat scheme was confusing -- it made post-hoc-assigned noise look like\nreal cluster membership -- and has been replaced by the honest `N-<n>`\nlabeling described above. If you have data keyed by the old `U-<n>` names,\ncross-reference via `legacy_name`, which is unchanged by any of this.)\n\nWithin a family (not for `N` solutions -- see below), every member is\nrotated/reflected/shifted to minimize its distance to the family's own\nreference solution (member 1), so browsing a family actually shows one\nrecurring shape, not 16 arbitrary orientations of it. The wave speed `c`\nis sign-corrected to stay consistent with this re-orientation. `N-<n>`\nsolutions are **not** aligned to anything -- there is no family to align\nto -- and are stored in their raw, as-solved orientation. `legacy_name`\nrecords the original identifier (e.g. `EQ7`, `TW118`) from before any of\nthis renaming, for anyone cross-referencing older material; it is stable\nacross every naming scheme this project has ever used.\n\n### Fields (`solutions.h5`)\n\n| Field | Shape | Description |\n|---|---|---|\n| `w` | (8952,128,128) | vorticity field, symmetry-aligned |\n| `c` | (8952,) | wave speed (0 for equilibria), sign-corrected to match `w` |\n| `x`, `y` | (128,128) | full coordinate grids |\n| `forcing` | (128,128) | -4cos(4y), same shape as `w`, ready to use directly |\n| `name` | (8952,) | primary name, e.g. `\"F17-3\"` or `\"N-42\"` |\n| `family` | (8952,) | -1 = N (genuine HDBSCAN noise), 1-416 = family number |\n| `solution` | (8952,","encoding":"utf8","key":"README.md","name":"README.md","size":7915,"truncated":true}],"sourceObjectKey":"solutions.h5","starterCell":"from huggingface_hub import hf_hub_download\npath = hf_hub_download(repo_id=\"mgolden30/Kolmogorov_TW\", filename=\"solutions.h5\", repo_type=\"dataset\")\nprint(path)","starterSamplePath":"solutions.h5","totalObjects":6},"summary":"A database of exact steady and relative-equilibrium solutions of 2D Kolmogorov flow, the incompressible Navier-Stokes equations driven by a sinusoidal body force, at the…","title":"Kolmogorov Flow Exact Coherent States: 8,952 Equilibria and Traveling Waves of 2D Navier-Stokes (nu = 1/40)","updatedAt":"2026-11-22T00:59:43.319Z","visibility":"public","websiteUrl":"https://huggingface.co/datasets/mgolden30/Kolmogorov_TW"}}