{"id":51650701,"url":"https://github.com/replikativ/evoc-rstr","last_synced_at":"2026-07-14T03:32:42.786Z","repository":{"id":369429868,"uuid":"1272791723","full_name":"replikativ/evoc-rstr","owner":"replikativ","description":"EVoC (Embedding Vector Oriented Clustering) for Clojure — a port of the Tutte Institute's evoc on raster + umap-rstr.","archived":false,"fork":false,"pushed_at":"2026-07-05T08:50:37.000Z","size":784,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-07-05T10:13:29.847Z","etag":null,"topics":["clojure","data-science","raster","umap","visualization"],"latest_commit_sha":null,"homepage":"https://simm.is","language":"Clojure","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"bsd-2-clause","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/replikativ.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":"NOTICE","maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-18T00:04:49.000Z","updated_at":"2026-07-05T08:50:41.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/replikativ/evoc-rstr","commit_stats":null,"previous_names":["replikativ/evoc-rstr"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/replikativ/evoc-rstr","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/replikativ%2Fevoc-rstr","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/replikativ%2Fevoc-rstr/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/replikativ%2Fevoc-rstr/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/replikativ%2Fevoc-rstr/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/replikativ","download_url":"https://codeload.github.com/replikativ/evoc-rstr/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/replikativ%2Fevoc-rstr/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35445235,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-14T02:00:06.603Z","response_time":114,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["clojure","data-science","raster","umap","visualization"],"created_at":"2026-07-14T03:32:42.252Z","updated_at":"2026-07-14T03:32:42.777Z","avatar_url":"https://github.com/replikativ.png","language":"Clojure","funding_links":[],"categories":[],"sub_categories":[],"readme":"# evoc-rstr\n\n[![Clojars Project](https://img.shields.io/clojars/v/org.replikativ/evoc-rstr.svg)](https://clojars.org/org.replikativ/evoc-rstr)\n[![CircleCI](https://circleci.com/gh/replikativ/evoc-rstr.svg?style=shield)](https://circleci.com/gh/replikativ/evoc-rstr)\n[![Slack](https://img.shields.io/badge/slack-join_chat-brightgreen.svg)](https://clojurians.slack.com/archives/C09622F337D)\n\nEVoC (Embedding Vector Oriented Clustering) for Clojure, built on\n[raster](https://github.com/replikativ/raster) +\n[umap-rstr](https://github.com/replikativ/umap-rstr). A port of the reference\n[`evoc`](https://github.com/TutteInstitute/evoc) (Tutte Institute / Leland\nMcInnes) — embed with a UMAP-style layout, then HDBSCAN\\*-family clustering with\nmulti-resolution persistence-based cluster selection. The numeric kernels are\n`deftm` functions that JIT-compile to primitive-speed JVM bytecode. The `-rstr`\nsuffix in the repo/artifact name marks the raster substrate (à la Julia's `.jl`);\nthe namespace you require is just `evoc`.\n\n## Status — functional\n\nFull pipeline ported and validated against the reference `evoc` 0.3.1: cosine kNN →\nfuzzy simplicial set → reproducible node embedding → HDBSCAN\\* spine (mutual-reach\nMST → condensed tree → labels) → **multi-resolution persistence layer selection** →\n`labels_ = argmax(persistence)`.\n\n**Every stage is bit-exact to the reference** on committed gold fixtures (per-stage\nregression tests, `test/evoc/gold_test.clj`): MST total weight; condense + labels\n(ARI 1.0 on the gold linkage); fuzzy-graph weights; embedding kernel (1 real epoch);\nmembership strengths; barcode / persistence / peaks / selected layers.\n\n**End-to-end** (EVoC's own test methodology — determinism + ARI, since EVoC is\ninherently seed-unstable, not exact-output): on MNIST n=4000, evoc-rstr finds 12\nclusters (14% noise) matching the reference's 12 (15%), **ARI(evoc-rstr, reference)\n= 0.67** (vs EVoC's cross-seed floor ~0.06), ARI-vs-truth 0.629 (reference 0.595).\nSee `dev/evoc_{gen.py,compare.clj,plot.py}` for the comparison harness + plot.\n\nClustering side-by-side on the same 2-D UMAP layout (left: true labels; middle:\nreference EVoC; right: evoc-rstr):\n\n![EVoC clustering: reference vs evoc-rstr on MNIST](img/evoc_compare.png)\n\n**Intentional deviation:** the embedding uses a deterministic random init\n(= EVoC's supported `node_embedding_init=None`). The default `label_propagation_init`\n(a recursive multilevel graph-coarsening init) is **not ported** — measured to give\nno clustering-quality advantage over random at these scales (it's a large-n\nconvergence/scalability technique), and it can't be bit-reproduced without numpy's\nRNG. It remains optional polish if exact default-seed reproduction is ever needed.\n\n## Usage\n\n```clojure\n(require '[evoc :as evoc])\n\n;; X: flat row-major double[] (or float[]) of n*dim\n(def result (evoc/fit-predict X n dim :k 15 :min-samples 5 :seed 42))\n(:labels result)      ;; =\u003e int[n] cluster labels (noise = -1)\n(:n-clusters result)  ;; =\u003e number of clusters found\n(:embedding result)   ;; =\u003e double[n*2] 2-D layout used for clustering\n```\n\nOptions: `:k` (neighbors, 15), `:n-epochs` (50), `:min-samples` (5),\n`:base-min-cluster-size` (5), `:noise` (0.5), `:max-layers` (10),\n`:min-similarity` (0.2), `:seed` (42). The result also carries `:persistence`,\n`:layers` (the multi-resolution candidate labelings), and `:strengths`.\n\n## Namespaces\n\n- `evoc.mst` — mutual-reachability MST (Prim + Boruvka switch)\n- `evoc.boruvka` — KD-tree Boruvka MST (large n)\n- `evoc.tree` — linkage / condensed tree / leaf extraction / labelling\n- `evoc.embed` — EVoC node-embedding SGD (reproducible deferred-update epoch)\n- `evoc.layers` — multi-resolution persistence layer selection (the defining EVoC feature)\n- `evoc` — `fit-predict` orchestrator\n\nkNN / KD-trees / RNG live in raster (`raster.knn`, `raster.spatial.*`); the fuzzy\nsimplicial set lives in umap-rstr (`umap.graph`).\n\n## Installation\n\nraster (and umap-rstr) depend on a typedclojure fork via git, and git deps\nresolve transitively only through `deps.edn` (not a Maven POM), so pin this lib\nas a git dependency:\n\n```clojure\nio.github.replikativ/evoc-rstr\n{:git/url \"https://github.com/replikativ/evoc-rstr\"\n :git/sha \"\u003csha\u003e\"}\n```\n\n(Once typedclojure ships its fixes in a Maven release upstream, the whole chain\nmoves to `org.replikativ/*` Clojars coordinates.)\n\n## Requirements\n\nValhalla JDK (raster's `deftm` kernels use preview features). Run with `:valhalla`.\n\n```bash\nclojure -M:valhalla:test\n```\n\n## License\n\nBSD 2-Clause (see `LICENSE` and `NOTICE`).\n\nevoc-rstr is a derivative work — a Clojure / raster port of\n[evoc](https://github.com/TutteInstitute/evoc) (© 2024 Tutte Institute for\nMathematics and Computing, BSD 2-Clause). It follows the reference evoc's\nalgorithm and numerical behaviour (validated bit-exact per-stage against evoc\n0.3.1) but is an independent reimplementation, not endorsed by or affiliated with\nthe original authors. The port is © 2026 Christian Weilbach, released under the\nsame license.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Freplikativ%2Fevoc-rstr","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Freplikativ%2Fevoc-rstr","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Freplikativ%2Fevoc-rstr/lists"}