prosemble 2.1.1__tar.gz → 2.1.2__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (288) hide show
  1. {prosemble-2.1.1 → prosemble-2.1.2}/.github/workflows/ci.yml +1 -1
  2. {prosemble-2.1.1 → prosemble-2.1.2}/PKG-INFO +1 -1
  3. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/distance.py +7 -12
  4. prosemble-2.1.2/prosemble/core/kernel.py +222 -0
  5. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/losses.py +5 -5
  6. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/similarities.py +2 -2
  7. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/__init__.py +15 -0
  8. prosemble-2.1.2/prosemble/models/dk_glvq.py +246 -0
  9. prosemble-2.1.2/prosemble/models/dk_heskes_som.py +230 -0
  10. prosemble-2.1.2/prosemble/models/dk_kohonen_som.py +219 -0
  11. prosemble-2.1.2/prosemble/models/dk_matrix_lvq.py +248 -0
  12. prosemble-2.1.2/prosemble/models/dk_neural_gas.py +174 -0
  13. prosemble-2.1.2/prosemble/models/dk_relevance_lvq.py +274 -0
  14. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble.egg-info/PKG-INFO +1 -1
  15. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble.egg-info/SOURCES.txt +8 -0
  16. {prosemble-2.1.1 → prosemble-2.1.2}/pyproject.toml +1 -1
  17. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/api/models.rst +33 -0
  18. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/conf.py +1 -1
  19. prosemble-2.1.2/sphinx-docs/guides/differentiating_kernels.rst +294 -0
  20. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/index.rst +1 -0
  21. prosemble-2.1.2/tests/test_dk_models.py +379 -0
  22. prosemble-2.1.1/prosemble/core/kernel.py +0 -103
  23. {prosemble-2.1.1 → prosemble-2.1.2}/.bumpversion.cfg +0 -0
  24. {prosemble-2.1.1 → prosemble-2.1.2}/.editorconfig +0 -0
  25. {prosemble-2.1.1 → prosemble-2.1.2}/.envrc +0 -0
  26. {prosemble-2.1.1 → prosemble-2.1.2}/.github/ISSUE_TEMPLATE/bug_report.md +0 -0
  27. {prosemble-2.1.1 → prosemble-2.1.2}/.github/ISSUE_TEMPLATE/config.yml +0 -0
  28. {prosemble-2.1.1 → prosemble-2.1.2}/.github/ISSUE_TEMPLATE/feature_request.md +0 -0
  29. {prosemble-2.1.1 → prosemble-2.1.2}/.github/workflows/release.yml +0 -0
  30. {prosemble-2.1.1 → prosemble-2.1.2}/.gitignore +0 -0
  31. {prosemble-2.1.1 → prosemble-2.1.2}/.readthedocs.yaml +0 -0
  32. {prosemble-2.1.1 → prosemble-2.1.2}/CONSTITUTION.md +0 -0
  33. {prosemble-2.1.1 → prosemble-2.1.2}/LICENSE +0 -0
  34. {prosemble-2.1.1 → prosemble-2.1.2}/MANIFEST.in +0 -0
  35. {prosemble-2.1.1 → prosemble-2.1.2}/ONNX_COVERAGE.md +0 -0
  36. {prosemble-2.1.1 → prosemble-2.1.2}/README.md +0 -0
  37. {prosemble-2.1.1 → prosemble-2.1.2}/TASK.md +0 -0
  38. {prosemble-2.1.1 → prosemble-2.1.2}/benchmarks/benchmark_distance_jax.py +0 -0
  39. {prosemble-2.1.1 → prosemble-2.1.2}/benchmarks/benchmark_fcm_jax.py +0 -0
  40. {prosemble-2.1.1 → prosemble-2.1.2}/benchmarks/benchmark_pcm_jax.py +0 -0
  41. {prosemble-2.1.1 → prosemble-2.1.2}/devenv.lock +0 -0
  42. {prosemble-2.1.1 → prosemble-2.1.2}/devenv.nix +0 -0
  43. {prosemble-2.1.1 → prosemble-2.1.2}/devenv.yaml +0 -0
  44. {prosemble-2.1.1 → prosemble-2.1.2}/examples/afcm_iris.py +0 -0
  45. {prosemble-2.1.1 → prosemble-2.1.2}/examples/bgpc_iris.py +0 -0
  46. {prosemble-2.1.1 → prosemble-2.1.2}/examples/cbc_iris.py +0 -0
  47. {prosemble-2.1.1 → prosemble-2.1.2}/examples/celvq_ng_iris.py +0 -0
  48. {prosemble-2.1.1 → prosemble-2.1.2}/examples/fcm_iris.py +0 -0
  49. {prosemble-2.1.1 → prosemble-2.1.2}/examples/fpcm_iris.py +0 -0
  50. {prosemble-2.1.1 → prosemble-2.1.2}/examples/glvq_advanced_iris.py +0 -0
  51. {prosemble-2.1.1 → prosemble-2.1.2}/examples/glvq_iris.py +0 -0
  52. {prosemble-2.1.1 → prosemble-2.1.2}/examples/glvq_unequal_prototypes_iris.py +0 -0
  53. {prosemble-2.1.1 → prosemble-2.1.2}/examples/gmlvq_iris.py +0 -0
  54. {prosemble-2.1.1 → prosemble-2.1.2}/examples/grlvq_iris.py +0 -0
  55. {prosemble-2.1.1 → prosemble-2.1.2}/examples/hcm_iris.py +0 -0
  56. {prosemble-2.1.1 → prosemble-2.1.2}/examples/heskes_som_iris.py +0 -0
  57. {prosemble-2.1.1 → prosemble-2.1.2}/examples/ipcm2_iris.py +0 -0
  58. {prosemble-2.1.1 → prosemble-2.1.2}/examples/ipcm_iris.py +0 -0
  59. {prosemble-2.1.1 → prosemble-2.1.2}/examples/kafcm_iris.py +0 -0
  60. {prosemble-2.1.1 → prosemble-2.1.2}/examples/kfcm_iris.py +0 -0
  61. {prosemble-2.1.1 → prosemble-2.1.2}/examples/kfpcm_iris.py +0 -0
  62. {prosemble-2.1.1 → prosemble-2.1.2}/examples/kipcm2_iris.py +0 -0
  63. {prosemble-2.1.1 → prosemble-2.1.2}/examples/kipcm_iris.py +0 -0
  64. {prosemble-2.1.1 → prosemble-2.1.2}/examples/kmeans_iris.py +0 -0
  65. {prosemble-2.1.1 → prosemble-2.1.2}/examples/knn_iris.py +0 -0
  66. {prosemble-2.1.1 → prosemble-2.1.2}/examples/kohonen_som_iris.py +0 -0
  67. {prosemble-2.1.1 → prosemble-2.1.2}/examples/kpcm_iris.py +0 -0
  68. {prosemble-2.1.1 → prosemble-2.1.2}/examples/kpfcm_iris.py +0 -0
  69. {prosemble-2.1.1 → prosemble-2.1.2}/examples/large_dataset_training.py +0 -0
  70. {prosemble-2.1.1 → prosemble-2.1.2}/examples/lcelvq_ng_iris.py +0 -0
  71. {prosemble-2.1.1 → prosemble-2.1.2}/examples/lmrslvq_iris.py +0 -0
  72. {prosemble-2.1.1 → prosemble-2.1.2}/examples/lmrslvq_ng_iris.py +0 -0
  73. {prosemble-2.1.1 → prosemble-2.1.2}/examples/lvq1_iris.py +0 -0
  74. {prosemble-2.1.1 → prosemble-2.1.2}/examples/lvqmln_iris.py +0 -0
  75. {prosemble-2.1.1 → prosemble-2.1.2}/examples/mcelvq_ng_iris.py +0 -0
  76. {prosemble-2.1.1 → prosemble-2.1.2}/examples/mrslvq_iris.py +0 -0
  77. {prosemble-2.1.1 → prosemble-2.1.2}/examples/mrslvq_ng_iris.py +0 -0
  78. {prosemble-2.1.1 → prosemble-2.1.2}/examples/neural_gas_iris.py +0 -0
  79. {prosemble-2.1.1 → prosemble-2.1.2}/examples/npc_iris.py +0 -0
  80. {prosemble-2.1.1 → prosemble-2.1.2}/examples/oc_lmrslvq_iris.py +0 -0
  81. {prosemble-2.1.1 → prosemble-2.1.2}/examples/oc_lmrslvq_ng_iris.py +0 -0
  82. {prosemble-2.1.1 → prosemble-2.1.2}/examples/oc_mrslvq_iris.py +0 -0
  83. {prosemble-2.1.1 → prosemble-2.1.2}/examples/oc_mrslvq_ng_iris.py +0 -0
  84. {prosemble-2.1.1 → prosemble-2.1.2}/examples/oc_rslvq_iris.py +0 -0
  85. {prosemble-2.1.1 → prosemble-2.1.2}/examples/oc_rslvq_ng_iris.py +0 -0
  86. {prosemble-2.1.1 → prosemble-2.1.2}/examples/pcm_iris.py +0 -0
  87. {prosemble-2.1.1 → prosemble-2.1.2}/examples/pfcm_iris.py +0 -0
  88. {prosemble-2.1.1 → prosemble-2.1.2}/examples/plvq_iris.py +0 -0
  89. {prosemble-2.1.1 → prosemble-2.1.2}/examples/rslvq_ng_iris.py +0 -0
  90. {prosemble-2.1.1 → prosemble-2.1.2}/examples/slng_iris.py +0 -0
  91. {prosemble-2.1.1 → prosemble-2.1.2}/examples/smng_iris.py +0 -0
  92. {prosemble-2.1.1 → prosemble-2.1.2}/examples/som_iris.py +0 -0
  93. {prosemble-2.1.1 → prosemble-2.1.2}/examples/srng_iris.py +0 -0
  94. {prosemble-2.1.1 → prosemble-2.1.2}/examples/stng_iris.py +0 -0
  95. {prosemble-2.1.1 → prosemble-2.1.2}/examples/tcelvq_ng_iris.py +0 -0
  96. {prosemble-2.1.1 → prosemble-2.1.2}/examples/vis_fcm_iris.py +0 -0
  97. {prosemble-2.1.1 → prosemble-2.1.2}/examples/vis_glvq_iris.py +0 -0
  98. {prosemble-2.1.1 → prosemble-2.1.2}/examples/vis_neural_gas_iris.py +0 -0
  99. {prosemble-2.1.1 → prosemble-2.1.2}/examples/vis_som_iris.py +0 -0
  100. {prosemble-2.1.1 → prosemble-2.1.2}/flake.nix +0 -0
  101. {prosemble-2.1.1 → prosemble-2.1.2}/mkdocs.yml +0 -0
  102. {prosemble-2.1.1 → prosemble-2.1.2}/nix/package.nix +0 -0
  103. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/__init__.py +0 -0
  104. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/__init__.py +0 -0
  105. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/activations.py +0 -0
  106. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/callbacks.py +0 -0
  107. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/competitions.py +0 -0
  108. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/data.py +0 -0
  109. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/distributed.py +0 -0
  110. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/initializers.py +0 -0
  111. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/manifolds.py +0 -0
  112. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/model_selection.py +0 -0
  113. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/onnx_export.py +0 -0
  114. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/pipeline.py +0 -0
  115. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/pooling.py +0 -0
  116. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/protocols.py +0 -0
  117. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/quantization.py +0 -0
  118. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/serialization.py +0 -0
  119. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/utils.py +0 -0
  120. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/vis.py +0 -0
  121. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/core/visualizer.py +0 -0
  122. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/datasets/__init__.py +0 -0
  123. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/datasets/dataset.py +0 -0
  124. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/afcm.py +0 -0
  125. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/base.py +0 -0
  126. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/bgpc.py +0 -0
  127. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/cbc.py +0 -0
  128. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/celvq_ng.py +0 -0
  129. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/celvq_ng_mixin.py +0 -0
  130. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/crossentropy_lvq.py +0 -0
  131. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/fcm.py +0 -0
  132. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/fpcm.py +0 -0
  133. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/glvq.py +0 -0
  134. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/growing_neural_gas.py +0 -0
  135. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/hcm.py +0 -0
  136. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/heskes_som.py +0 -0
  137. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/image_cbc.py +0 -0
  138. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/image_lvq.py +0 -0
  139. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/ipcm.py +0 -0
  140. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/ipcm2.py +0 -0
  141. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/kafcm.py +0 -0
  142. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/kfcm.py +0 -0
  143. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/kfpcm.py +0 -0
  144. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/kipcm.py +0 -0
  145. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/kipcm2.py +0 -0
  146. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/kmeans.py +0 -0
  147. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/knn.py +0 -0
  148. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/kohonen_som.py +0 -0
  149. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/kpcm.py +0 -0
  150. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/kpfcm.py +0 -0
  151. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/lcelvq_ng.py +0 -0
  152. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/local_matrix_lvq.py +0 -0
  153. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/lvq1.py +0 -0
  154. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/lvq21.py +0 -0
  155. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/lvqmln.py +0 -0
  156. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/matrix_lvq.py +0 -0
  157. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/matrix_rslvq.py +0 -0
  158. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/mcelvq_ng.py +0 -0
  159. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/median_lvq.py +0 -0
  160. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/mrslvq_ng.py +0 -0
  161. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/neural_gas.py +0 -0
  162. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/npc.py +0 -0
  163. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_glvq.py +0 -0
  164. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_glvq_ng.py +0 -0
  165. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_glvq_ng_mixin.py +0 -0
  166. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_gmlvq.py +0 -0
  167. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_gmlvq_ng.py +0 -0
  168. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_grlvq.py +0 -0
  169. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_grlvq_ng.py +0 -0
  170. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_gtlvq.py +0 -0
  171. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_gtlvq_ng.py +0 -0
  172. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_lgmlvq.py +0 -0
  173. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_lgmlvq_ng.py +0 -0
  174. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_mrslvq.py +0 -0
  175. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_mrslvq_ng.py +0 -0
  176. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_rslvq.py +0 -0
  177. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_rslvq_ng.py +0 -0
  178. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/oc_rslvq_ng_mixin.py +0 -0
  179. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/pcm.py +0 -0
  180. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/pfcm.py +0 -0
  181. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/plvq.py +0 -0
  182. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/probabilistic_lvq.py +0 -0
  183. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/prototype_base.py +0 -0
  184. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/relevance_lvq.py +0 -0
  185. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/riemannian_neural_gas.py +0 -0
  186. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/riemannian_slng.py +0 -0
  187. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/riemannian_smng.py +0 -0
  188. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/riemannian_srng.py +0 -0
  189. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/riemannian_stng.py +0 -0
  190. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/rslvq_ng.py +0 -0
  191. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/siamese_lvq.py +0 -0
  192. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/slng.py +0 -0
  193. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/smng.py +0 -0
  194. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/som.py +0 -0
  195. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/srng.py +0 -0
  196. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/stng.py +0 -0
  197. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/svq_occ.py +0 -0
  198. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/svq_occ_lm.py +0 -0
  199. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/svq_occ_m.py +0 -0
  200. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/svq_occ_r.py +0 -0
  201. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/svq_occ_t.py +0 -0
  202. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/tangent_lvq.py +0 -0
  203. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/models/tcelvq_ng.py +0 -0
  204. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble/py.typed +0 -0
  205. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble.egg-info/dependency_links.txt +0 -0
  206. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble.egg-info/requires.txt +0 -0
  207. {prosemble-2.1.1 → prosemble-2.1.2}/prosemble.egg-info/top_level.txt +0 -0
  208. {prosemble-2.1.1 → prosemble-2.1.2}/requirements_docs.txt +0 -0
  209. {prosemble-2.1.1 → prosemble-2.1.2}/setup.cfg +0 -0
  210. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/Makefile +0 -0
  211. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/api/core.rst +0 -0
  212. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/guides/advanced.rst +0 -0
  213. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/guides/clustering.rst +0 -0
  214. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/guides/one_class.rst +0 -0
  215. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/guides/onnx.rst +0 -0
  216. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/guides/riemannian.rst +0 -0
  217. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/guides/supervised.rst +0 -0
  218. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/guides/topology.rst +0 -0
  219. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/installation.rst +0 -0
  220. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/make.bat +0 -0
  221. {prosemble-2.1.1 → prosemble-2.1.2}/sphinx-docs/quickstart.rst +0 -0
  222. {prosemble-2.1.1 → prosemble-2.1.2}/tests/__init__.py +0 -0
  223. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_activations.py +0 -0
  224. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_cbc.py +0 -0
  225. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_celvq_ng.py +0 -0
  226. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_checkpoint.py +0 -0
  227. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_clustering_patience.py +0 -0
  228. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_competitions.py +0 -0
  229. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_core.py +0 -0
  230. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_core_jax.py +0 -0
  231. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_data.py +0 -0
  232. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_dataset.py +0 -0
  233. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_dataset_jax.py +0 -0
  234. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_distance_jax.py +0 -0
  235. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_distributed.py +0 -0
  236. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_examples.sh +0 -0
  237. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_fcm_jax.py +0 -0
  238. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_glvq.py +0 -0
  239. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_gmlvq.py +0 -0
  240. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_heskes_som.py +0 -0
  241. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_image_lvq.py +0 -0
  242. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_initializers.py +0 -0
  243. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_jax_models.py +0 -0
  244. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_jit_inference.py +0 -0
  245. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_kohonen_som.py +0 -0
  246. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_lcelvq_ng.py +0 -0
  247. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_losses.py +0 -0
  248. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_lvq_nongradient.py +0 -0
  249. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_lvqmln_plvq.py +0 -0
  250. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_matrix_rslvq.py +0 -0
  251. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_mcelvq_ng.py +0 -0
  252. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_minibatch.py +0 -0
  253. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_mixed_precision.py +0 -0
  254. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_model_selection.py +0 -0
  255. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_mrslvq_ng.py +0 -0
  256. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_neural_gas.py +0 -0
  257. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_oc_glvq.py +0 -0
  258. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_oc_glvq_ng.py +0 -0
  259. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_oc_glvq_variants.py +0 -0
  260. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_oc_mrslvq.py +0 -0
  261. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_oc_rslvq.py +0 -0
  262. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_oc_rslvq_ng.py +0 -0
  263. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_onnx_export.py +0 -0
  264. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_pcm_jax.py +0 -0
  265. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_pfcm_jax.py +0 -0
  266. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_pipeline.py +0 -0
  267. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_pooling.py +0 -0
  268. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_probabilistic.py +0 -0
  269. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_protocols.py +0 -0
  270. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_prototype_base.py +0 -0
  271. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_quantization.py +0 -0
  272. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_riemannian_neural_gas.py +0 -0
  273. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_riemannian_supervised_ng.py +0 -0
  274. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_rslvq_ng.py +0 -0
  275. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_scan_training.py +0 -0
  276. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_serialization.py +0 -0
  277. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_siamese_lvq.py +0 -0
  278. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_similarities.py +0 -0
  279. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_slng.py +0 -0
  280. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_smng.py +0 -0
  281. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_srng.py +0 -0
  282. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_stng.py +0 -0
  283. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_svq_occ.py +0 -0
  284. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_svq_occ_r.py +0 -0
  285. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_svq_occ_variants.py +0 -0
  286. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_tcelvq_ng.py +0 -0
  287. {prosemble-2.1.1 → prosemble-2.1.2}/tests/test_vis.py +0 -0
  288. {prosemble-2.1.1 → prosemble-2.1.2}/uv.lock +0 -0
@@ -33,7 +33,7 @@ jobs:
33
33
  python -m pip install --upgrade pip
34
34
  pip install .[all]
35
35
  - name: Run tests
36
- run: pytest tests/ -x -q --tb=short --ignore=tests/test_oc_glvq.py --ignore=tests/test_oc_glvq_ng.py --ignore=tests/test_oc_glvq_variants.py --ignore=tests/test_svq_occ.py --ignore=tests/test_svq_occ_r.py --ignore=tests/test_svq_occ_variants.py --ignore=tests/test_smng.py --ignore=tests/test_slng.py --ignore=tests/test_stng.py --ignore=tests/test_riemannian_neural_gas.py
36
+ run: pytest tests/ -x -q --tb=short --ignore=tests/test_oc_glvq.py --ignore=tests/test_oc_glvq_ng.py --ignore=tests/test_oc_glvq_variants.py --ignore=tests/test_svq_occ.py --ignore=tests/test_svq_occ_r.py --ignore=tests/test_svq_occ_variants.py --ignore=tests/test_smng.py --ignore=tests/test_slng.py --ignore=tests/test_stng.py --ignore=tests/test_riemannian_neural_gas.py --ignore=tests/test_riemannian_supervised_ng.py
37
37
 
38
38
  build:
39
39
  name: Build package
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: prosemble
3
- Version: 2.1.1
3
+ Version: 2.1.2
4
4
  Summary: A python project for prototype-based machine learning models
5
5
  Author-email: Nana Abeka Otoo <nanaabekaotoo@gmail.com>
6
6
  License: MIT
@@ -380,17 +380,12 @@ def tangent_distance_matrix(
380
380
  """
381
381
  Compute pairwise localized tangent distances.
382
382
 
383
- Each prototype j has an orthogonal subspace basis Omega_j of shape (d, s).
383
+ Each prototype j has an orthogonal subspace basis :math:`\Omega_j` of shape (d, s).
384
384
  The tangent distance projects out the subspace directions:
385
385
 
386
- d(x, w_j) = ||(I - Omega_j @ Omega_j^T)(x - w_j)||^2
386
+ .. math::
387
387
 
388
- This is equivalent to:
389
- diff = x - w_j
390
- proj = Omega_j^T @ diff (project onto subspace)
391
- recon = Omega_j @ proj (reconstruct in ambient space)
392
- tangent_diff = diff - recon (residual orthogonal to subspace)
393
- d = ||tangent_diff||^2
388
+ d(x, w_j) = \|(I - \Omega_j \Omega_j^T)(x - w_j)\|^2
394
389
 
395
390
  Parameters
396
391
  ----------
@@ -484,9 +479,9 @@ def gaussian_kernel_matrix(
484
479
  True
485
480
 
486
481
  Kernel Trick:
487
- For feature map phi mapping to infinite-dimensional Hilbert space,
488
- K(x, y) = <phi(x), phi(y)>. Kernel distance:
489
- ||phi(x) - phi(y)||^2 = K(x,x) - 2K(x,y) + K(y,y) = 2 - 2K(x,y) for normalized kernel.
482
+ For feature map :math:`\phi` mapping to infinite-dimensional Hilbert space,
483
+ :math:`K(x, y) = \langle\phi(x), \phi(y)\rangle`. Kernel distance:
484
+ :math:`\|\phi(x) - \phi(y)\|^2 = K(x,x) - 2K(x,y) + K(y,y) = 2 - 2K(x,y)` for normalized kernel.
490
485
 
491
486
  Use Cases:
492
487
  - Kernel Fuzzy C-Means (KFCM)
@@ -643,7 +638,7 @@ def omega_distance(x: chex.Array, y: chex.Array, omega: chex.Array) -> chex.Arra
643
638
  """
644
639
  Omega (projection-based) distance between two vectors.
645
640
 
646
- Computes ||diff @ omega||² where diff = x - y.
641
+ Computes :math:`\|\text{diff} \cdot \Omega\|^2` where :math:`\text{diff} = x - y`.
647
642
 
648
643
  Args:
649
644
  x: Array of shape (d,)
@@ -0,0 +1,222 @@
1
+ """
2
+ JAX-based kernel functions for kernel clustering methods.
3
+
4
+ This module provides GPU-accelerated kernel computations using JAX.
5
+ """
6
+
7
+ import jax.numpy as jnp
8
+ import chex
9
+ from jax import jit
10
+ from functools import partial
11
+
12
+
13
+ @jit
14
+ def gaussian_kernel(x: chex.Array, y: chex.Array, sigma: float) -> chex.Array:
15
+ """
16
+ Compute Gaussian (RBF) kernel between two vectors.
17
+
18
+ :math:`K(x, y) = \exp(-\|x - y\|^2 / (2\sigma^2))`
19
+
20
+ Args:
21
+ x: First vector, shape (n_features,)
22
+ y: Second vector, shape (n_features,)
23
+ sigma: Kernel bandwidth parameter
24
+
25
+ Returns:
26
+ Kernel value (scalar)
27
+ """
28
+ diff = x - y
29
+ sq_norm = jnp.sum(diff * diff)
30
+ return jnp.exp(-sq_norm / (2.0 * sigma ** 2))
31
+
32
+
33
+ @jit
34
+ def batch_gaussian_kernel(
35
+ X: chex.Array, Y: chex.Array, sigma: float
36
+ ) -> chex.Array:
37
+ """
38
+ Compute Gaussian kernel between two sets of vectors.
39
+
40
+ Args:
41
+ X: First set of vectors, shape (n_samples, n_features)
42
+ Y: Second set of vectors, shape (m_samples, n_features)
43
+ sigma: Kernel bandwidth parameter
44
+
45
+ Returns:
46
+ Kernel matrix, shape (n_samples, m_samples)
47
+ K[i, j] = K(X[i], Y[j])
48
+ """
49
+ # Compute squared Euclidean distances
50
+ X_sq = jnp.sum(X ** 2, axis=1, keepdims=True) # (n_samples, 1)
51
+ Y_sq = jnp.sum(Y ** 2, axis=1, keepdims=True) # (m_samples, 1)
52
+
53
+ # ||x - y||^2 = ||x||^2 + ||y||^2 - 2<x, y>
54
+ sq_distances = X_sq + Y_sq.T - 2.0 * (X @ Y.T) # (n_samples, m_samples)
55
+
56
+ # Ensure non-negative (numerical stability)
57
+ sq_distances = jnp.maximum(sq_distances, 0.0)
58
+
59
+ # Compute kernel
60
+ K = jnp.exp(-sq_distances / (2.0 * sigma ** 2))
61
+
62
+ return K
63
+
64
+
65
+ @jit
66
+ def kernel_distance_squared(
67
+ X: chex.Array, Y: chex.Array, sigma: float
68
+ ) -> chex.Array:
69
+ """
70
+ Compute squared distance in feature space.
71
+
72
+ For Gaussian kernel: :math:`\|\phi(x) - \phi(y)\|^2 = K(x,x) + K(y,y) - 2K(x,y) = 2(1 - K(x,y))`,
73
+ since :math:`K(x,x) = 1`.
74
+
75
+ Args:
76
+ X: First set of vectors, shape (n_samples, n_features)
77
+ Y: Second set of vectors, shape (m_samples, n_features)
78
+ sigma: Kernel bandwidth parameter
79
+
80
+ Returns:
81
+ Squared distances in feature space, shape (n_samples, m_samples)
82
+ """
83
+ K = batch_gaussian_kernel(X, Y, sigma)
84
+ return 2.0 * (1.0 - K)
85
+
86
+
87
+ @jit
88
+ def kernel_distance(
89
+ X: chex.Array, Y: chex.Array, sigma: float
90
+ ) -> chex.Array:
91
+ """
92
+ Compute distance in feature space.
93
+
94
+ Args:
95
+ X: First set of vectors, shape (n_samples, n_features)
96
+ Y: Second set of vectors, shape (m_samples, n_features)
97
+ sigma: Kernel bandwidth parameter
98
+
99
+ Returns:
100
+ Distances in feature space, shape (n_samples, m_samples)
101
+ """
102
+ D_sq = kernel_distance_squared(X, Y, sigma)
103
+ return jnp.sqrt(jnp.maximum(D_sq, 0.0))
104
+
105
+
106
+ @jit
107
+ def kernel_distance_squared_per_proto(
108
+ X: chex.Array, W: chex.Array, sigmas: chex.Array
109
+ ) -> chex.Array:
110
+ """
111
+ Squared kernel distance with per-prototype bandwidth.
112
+
113
+ .. math::
114
+
115
+ d_\\kappa^2(x, w_k) = 2\\left(1 - \\exp\\left(-\\frac{\\|x - w_k\\|^2}{2\\sigma_k^2}\\right)\\right)
116
+
117
+ Each prototype :math:`w_k` has its own bandwidth :math:`\\sigma_k`.
118
+
119
+ Args:
120
+ X: Data matrix, shape (n_samples, n_features).
121
+ W: Prototype matrix, shape (n_prototypes, n_features).
122
+ sigmas: Per-prototype bandwidths, shape (n_prototypes,).
123
+
124
+ Returns:
125
+ Squared distances in feature space, shape (n_samples, n_prototypes).
126
+
127
+ References
128
+ ----------
129
+ .. [1] Villmann, T., Haase, S., & Kaden, M. (2015). Kernelized vector
130
+ quantization in gradient-descent learning. Neurocomputing.
131
+ """
132
+ diff = X[:, None, :] - W[None, :, :] # (n, p, d)
133
+ sq_norms = jnp.sum(diff ** 2, axis=2) # (n, p)
134
+ K = jnp.exp(-sq_norms / (2.0 * sigmas[None, :] ** 2)) # (n, p)
135
+ return 2.0 * (1.0 - K)
136
+
137
+
138
+ @jit
139
+ def kernel_distance_squared_relevance(
140
+ X: chex.Array, W: chex.Array, sigmas: chex.Array,
141
+ relevances: chex.Array
142
+ ) -> chex.Array:
143
+ """
144
+ Squared kernel distance with relevance weighting and per-prototype bandwidth.
145
+
146
+ .. math::
147
+
148
+ d_\\kappa^2(x, w_k) = 2\\left(1 - \\exp\\left(-\\frac{\\sum_j \\lambda_j (x_j - w_{kj})^2}{2\\sigma_k^2}\\right)\\right)
149
+
150
+ Args:
151
+ X: Data matrix, shape (n_samples, n_features).
152
+ W: Prototype matrix, shape (n_prototypes, n_features).
153
+ sigmas: Per-prototype bandwidths, shape (n_prototypes,).
154
+ relevances: Normalized relevance weights, shape (n_features,).
155
+
156
+ Returns:
157
+ Squared distances in feature space, shape (n_samples, n_prototypes).
158
+
159
+ References
160
+ ----------
161
+ .. [1] Villmann, T., Haase, S., & Kaden, M. (2015). Kernelized vector
162
+ quantization in gradient-descent learning. Neurocomputing.
163
+ """
164
+ diff = X[:, None, :] - W[None, :, :] # (n, p, d)
165
+ weighted_sq = jnp.sum(relevances[None, None, :] * diff ** 2, axis=2) # (n, p)
166
+ K = jnp.exp(-weighted_sq / (2.0 * sigmas[None, :] ** 2)) # (n, p)
167
+ return 2.0 * (1.0 - K)
168
+
169
+
170
+ @jit
171
+ def exponential_kernel_distance_squared(
172
+ X: chex.Array, W: chex.Array, omega_hat: chex.Array
173
+ ) -> chex.Array:
174
+ """
175
+ Squared distance in exponential kernel feature space.
176
+
177
+ Uses the exponential kernel :math:`\\kappa_{\\exp}(v, w, \\hat\\Lambda) = \\exp(v^T \\hat\\Lambda w)`
178
+ where :math:`\\hat\\Lambda = \\hat\\Omega \\hat\\Omega^T`.
179
+
180
+ .. math::
181
+
182
+ d_\\kappa^2(x, w) = \\exp(x^T \\hat\\Lambda x)
183
+ + \\exp(w^T \\hat\\Lambda w)
184
+ - 2 \\exp(x^T \\hat\\Lambda w)
185
+
186
+ Note: :math:`\\kappa(v, v) \\neq 1` for the exponential kernel, so the
187
+ full three-term formula is required (not the 2(1-K) simplification).
188
+
189
+ Args:
190
+ X: Data matrix, shape (n_samples, n_features).
191
+ W: Prototype matrix, shape (n_prototypes, n_features).
192
+ omega_hat: Transformation matrix, shape (n_features, latent_dim).
193
+ The kernel matrix is :math:`\\hat\\Lambda = \\hat\\Omega \\hat\\Omega^T`.
194
+
195
+ Returns:
196
+ Squared distances in feature space, shape (n_samples, n_prototypes).
197
+
198
+ References
199
+ ----------
200
+ .. [1] Villmann, T., Haase, S., & Kaden, M. (2015). Kernelized vector
201
+ quantization in gradient-descent learning. Neurocomputing.
202
+ """
203
+ # Λ̂ = Ω̂ Ω̂^T (d, d)
204
+ lambda_hat = jnp.dot(omega_hat, omega_hat.T)
205
+
206
+ # x^T Λ̂ x for all samples: (n,)
207
+ Lx = jnp.dot(X, lambda_hat) # (n, d)
208
+ xLx = jnp.sum(X * Lx, axis=1) # (n,)
209
+
210
+ # w^T Λ̂ w for all prototypes: (p,)
211
+ Lw = jnp.dot(W, lambda_hat) # (p, d)
212
+ wLw = jnp.sum(W * Lw, axis=1) # (p,)
213
+
214
+ # x^T Λ̂ w for all (n, p) pairs: (n, p)
215
+ xLw = jnp.dot(Lx, W.T) # (n, p)
216
+
217
+ # d_κ²(x, w) = exp(x^T Λ̂ x) + exp(w^T Λ̂ w) - 2·exp(x^T Λ̂ w)
218
+ distances = (jnp.exp(xLx[:, None])
219
+ + jnp.exp(wLw[None, :])
220
+ - 2.0 * jnp.exp(xLw))
221
+
222
+ return jnp.maximum(distances, 0.0)
@@ -74,7 +74,7 @@ def glvq_loss(distances, target_labels, prototype_labels,
74
74
  margin=0.0):
75
75
  """Generalized LVQ loss.
76
76
 
77
- mu_i = (d+_i - d-_i) / (d+_i + d-_i)
77
+ :math:`\mu_i = \frac{d^+_i - d^-_i}{d^+_i + d^-_i}`
78
78
 
79
79
  Parameters
80
80
  ----------
@@ -98,7 +98,7 @@ def glvq_loss_with_transfer(distances, target_labels, prototype_labels,
98
98
  transfer_fn=identity, margin=0.0, beta=10.0):
99
99
  """GLVQ loss with configurable transfer function.
100
100
 
101
- loss = mean(transfer(mu + margin, beta))
101
+ :math:`\text{loss} = \text{mean}(f(\mu + \text{margin}, \beta))`
102
102
 
103
103
  Parameters
104
104
  ----------
@@ -198,7 +198,7 @@ def _class_probabilities(distances, target_labels, prototype_labels, sigma):
198
198
  def nllr_loss(distances, target_labels, prototype_labels, sigma=1.0):
199
199
  """Negative Log-Likelihood Ratio loss (for SLVQ).
200
200
 
201
- loss = -log(P(correct) / P(wrong))
201
+ :math:`\text{loss} = -\log(P(\text{correct}) / P(\text{wrong}))`
202
202
 
203
203
  Parameters
204
204
  ----------
@@ -223,7 +223,7 @@ def nllr_loss(distances, target_labels, prototype_labels, sigma=1.0):
223
223
  def rslvq_loss(distances, target_labels, prototype_labels, sigma=1.0):
224
224
  """Robust Soft LVQ loss (for RSLVQ).
225
225
 
226
- loss = -log(P(correct) / P(all))
226
+ :math:`\text{loss} = -\log(P(\text{correct}) / P(\text{all}))`
227
227
 
228
228
  Parameters
229
229
  ----------
@@ -336,7 +336,7 @@ def cross_entropy_lvq_loss(distances, target_labels, prototype_labels, n_classes
336
336
  def margin_loss(y_pred, y_true_one_hot, margin=0.3):
337
337
  """Margin loss for CBC.
338
338
 
339
- loss = relu(max_wrong - correct + margin)
339
+ :math:`\text{loss} = \text{ReLU}(\max(\text{wrong}) - \text{correct} + \text{margin})`
340
340
 
341
341
  Parameters
342
342
  ----------
@@ -13,7 +13,7 @@ from jax import jit
13
13
  def gaussian_similarity(distances_sq, variance=1.0):
14
14
  """Convert squared distances to Gaussian similarities.
15
15
 
16
- s(d) = exp(-d^2 / (2 * variance))
16
+ :math:`s(d) = \exp(-d^2 / (2 \cdot \text{variance}))`
17
17
 
18
18
  Parameters
19
19
  ----------
@@ -34,7 +34,7 @@ def gaussian_similarity(distances_sq, variance=1.0):
34
34
  def cosine_similarity_matrix(X, Y):
35
35
  """Pairwise cosine similarity between rows of X and Y.
36
36
 
37
- cos(x, y) = (x . y) / (||x|| * ||y||)
37
+ :math:`\cos(x, y) = \frac{x \cdot y}{\|x\| \cdot \|y\|}`
38
38
 
39
39
  Parameters
40
40
  ----------
@@ -92,6 +92,14 @@ try:
92
92
  from .heskes_som import HeskesSOM
93
93
  from .riemannian_neural_gas import RiemannianNeuralGas
94
94
 
95
+ # Differentiating Kernel models (Villmann, Haase & Kaden, 2015)
96
+ from .dk_glvq import DKGLVQ
97
+ from .dk_relevance_lvq import DKGRLVQ
98
+ from .dk_matrix_lvq import DKGMLVQ
99
+ from .dk_neural_gas import DKNeuralGas
100
+ from .dk_kohonen_som import DKKohonenSOM
101
+ from .dk_heskes_som import DKHeskesSOM
102
+
95
103
  _MODEL_REGISTRY = {
96
104
  # Unsupervised clustering
97
105
  'FCM': FCM, 'PCM': PCM, 'PFCM': PFCM, 'AFCM': AFCM,
@@ -134,6 +142,10 @@ try:
134
142
  'NeuralGas': NeuralGas, 'GrowingNeuralGas': GrowingNeuralGas,
135
143
  'KohonenSOM': KohonenSOM, 'HeskesSOM': HeskesSOM,
136
144
  'RiemannianNeuralGas': RiemannianNeuralGas,
145
+ # Differentiating Kernel
146
+ 'DKGLVQ': DKGLVQ, 'DKGRLVQ': DKGRLVQ, 'DKGMLVQ': DKGMLVQ,
147
+ 'DKNeuralGas': DKNeuralGas, 'DKKohonenSOM': DKKohonenSOM,
148
+ 'DKHeskesSOM': DKHeskesSOM,
137
149
  }
138
150
 
139
151
  JAX_AVAILABLE = True
@@ -180,6 +192,9 @@ __all__ = [
180
192
  # Unsupervised topology
181
193
  'NeuralGas', 'GrowingNeuralGas', 'KohonenSOM', 'HeskesSOM',
182
194
  'RiemannianNeuralGas',
195
+ # Differentiating Kernel
196
+ 'DKGLVQ', 'DKGRLVQ', 'DKGMLVQ',
197
+ 'DKNeuralGas', 'DKKohonenSOM', 'DKHeskesSOM',
183
198
  # Status
184
199
  'JAX_AVAILABLE',
185
200
  ]
@@ -0,0 +1,246 @@
1
+ """
2
+ Differentiating Kernel GLVQ (DKGLVQ).
3
+
4
+ GLVQ with Gaussian kernel distance and per-prototype bandwidth adaptation.
5
+ Each prototype :math:`w_k` has a learnable bandwidth :math:`\\sigma_k`:
6
+
7
+ .. math::
8
+
9
+ d_\\kappa^2(x, w_k) = 2\\left(1 - \\exp\\left(
10
+ -\\frac{\\|x - w_k\\|^2}{2\\sigma_k^2}
11
+ \\right)\\right)
12
+
13
+ References
14
+ ----------
15
+ .. [1] Villmann, T., Haase, S., & Kaden, M. (2015). Kernelized vector
16
+ quantization in gradient-descent learning. Neurocomputing.
17
+ """
18
+
19
+ import jax
20
+ import jax.numpy as jnp
21
+ from jax import jit
22
+ import numpy as np
23
+
24
+ from prosemble.models.prototype_base import SupervisedPrototypeModel
25
+ from prosemble.core.competitions import wtac
26
+ from prosemble.core.kernel import kernel_distance_squared_per_proto
27
+
28
+
29
+ @jit
30
+ def _predict_dkglvq_jit(X, prototypes, sigmas, proto_labels, sigma_min):
31
+ """JIT-compiled DKGLVQ prediction with per-prototype kernel distance."""
32
+ sigmas = jnp.maximum(sigmas, sigma_min)
33
+ distances = kernel_distance_squared_per_proto(X, prototypes, sigmas)
34
+ return wtac(distances, proto_labels)
35
+
36
+
37
+ class DKGLVQ(SupervisedPrototypeModel):
38
+ """Differentiating Kernel GLVQ.
39
+
40
+ GLVQ with Gaussian kernel distance in feature space. Each prototype
41
+ :math:`w_k` has a learnable bandwidth :math:`\\sigma_k` that is
42
+ adapted via gradient descent alongside the prototypes.
43
+
44
+ .. math::
45
+
46
+ d_\\kappa^2(x, w_k) = 2\\left(1 - \\exp\\left(
47
+ -\\frac{\\|x - w_k\\|^2}{2\\sigma_k^2}
48
+ \\right)\\right)
49
+
50
+ Parameters
51
+ ----------
52
+ sigma_init : str or float
53
+ Initialization strategy for per-prototype bandwidths.
54
+ 'median' (default): per-class median distance from prototype to class members.
55
+ 'mean': per-class mean distance.
56
+ float: fixed value for all prototypes.
57
+ sigma_min : float
58
+ Lower bound for sigma to prevent bandwidth collapse. Default: 1e-3.
59
+ beta : float
60
+ Transfer function steepness parameter.
61
+ n_prototypes_per_class : int
62
+ Number of prototypes per class.
63
+ max_iter : int
64
+ Maximum training iterations.
65
+ lr : float
66
+ Learning rate.
67
+ epsilon : float
68
+ Convergence threshold on loss change.
69
+ random_seed : int
70
+ Random seed for reproducibility.
71
+ optimizer : str or optax optimizer, optional
72
+ Optimizer name ('adam', 'sgd') or optax GradientTransformation.
73
+ transfer_fn : callable, optional
74
+ Transfer function for loss shaping (default: identity).
75
+ margin : float
76
+ Margin for loss computation.
77
+ callbacks : list, optional
78
+ List of Callback objects.
79
+ use_scan : bool
80
+ If True (default), use jax.lax.scan for training.
81
+ batch_size : int, optional
82
+ Mini-batch size. If None, use full-batch training.
83
+ lr_scheduler : str or optax.Schedule, optional
84
+ Learning rate schedule.
85
+ lr_scheduler_kwargs : dict, optional
86
+ Keyword arguments for the learning rate scheduler.
87
+ prototypes_initializer : str or callable, optional
88
+ How to initialize prototypes.
89
+ patience : int, optional
90
+ Epochs with no improvement before stopping.
91
+ restore_best : bool
92
+ If True, restore best parameters after training.
93
+ class_weight : dict or 'balanced', optional
94
+ Weights for each class.
95
+ gradient_accumulation_steps : int, optional
96
+ Accumulate gradients over this many steps.
97
+ ema_decay : float, optional
98
+ Exponential moving average decay for parameters.
99
+ freeze_params : list of str, optional
100
+ Parameter group names to freeze.
101
+ lookahead : dict, optional
102
+ Lookahead optimizer wrapper configuration.
103
+ mixed_precision : str or None, optional
104
+ Compute dtype for mixed precision training.
105
+
106
+ References
107
+ ----------
108
+ .. [1] Villmann, T., Haase, S., & Kaden, M. (2015). Kernelized vector
109
+ quantization in gradient-descent learning. Neurocomputing.
110
+
111
+ See Also
112
+ --------
113
+ SupervisedPrototypeModel : Full list of base parameters.
114
+ """
115
+
116
+ def __init__(self, sigma_init='median', sigma_min=1e-3, beta=10.0,
117
+ n_prototypes_per_class=1, max_iter=100,
118
+ lr=0.01, epsilon=1e-6, random_seed=42, distance_fn=None,
119
+ optimizer='adam', transfer_fn=None, margin=0.0,
120
+ callbacks=None, use_scan=True, batch_size=None,
121
+ lr_scheduler=None, lr_scheduler_kwargs=None,
122
+ prototypes_initializer=None, patience=None,
123
+ restore_best=False, class_weight=None,
124
+ gradient_accumulation_steps=None, ema_decay=None,
125
+ freeze_params=None, lookahead=None, mixed_precision=None,
126
+ **kwargs):
127
+ super().__init__(
128
+ n_prototypes_per_class=n_prototypes_per_class,
129
+ max_iter=max_iter, lr=lr, epsilon=epsilon,
130
+ random_seed=random_seed, distance_fn=distance_fn,
131
+ optimizer=optimizer, transfer_fn=transfer_fn, margin=margin,
132
+ callbacks=callbacks, use_scan=use_scan, batch_size=batch_size,
133
+ lr_scheduler=lr_scheduler,
134
+ lr_scheduler_kwargs=lr_scheduler_kwargs,
135
+ prototypes_initializer=prototypes_initializer,
136
+ patience=patience, restore_best=restore_best,
137
+ class_weight=class_weight,
138
+ gradient_accumulation_steps=gradient_accumulation_steps,
139
+ ema_decay=ema_decay, freeze_params=freeze_params,
140
+ lookahead=lookahead, mixed_precision=mixed_precision,
141
+ **kwargs,
142
+ )
143
+ self.sigma_init = sigma_init
144
+ self.sigma_min = sigma_min
145
+ self.beta = beta
146
+ self.sigmas_ = None
147
+
148
+ def _estimate_sigmas(self, X, y, prototypes, proto_labels):
149
+ """Estimate per-prototype bandwidths from data."""
150
+ if isinstance(self.sigma_init, (int, float)):
151
+ return jnp.full(prototypes.shape[0], float(self.sigma_init))
152
+
153
+ sigmas = []
154
+ for k in range(prototypes.shape[0]):
155
+ label_k = proto_labels[k]
156
+ class_mask = (y == label_k)
157
+ X_class = X[class_mask]
158
+ dists = jnp.sqrt(jnp.sum((X_class - prototypes[k]) ** 2, axis=1))
159
+ if self.sigma_init == 'median':
160
+ sigma_k = jnp.median(dists)
161
+ else: # 'mean'
162
+ sigma_k = jnp.mean(dists)
163
+ sigmas.append(jnp.maximum(sigma_k, self.sigma_min))
164
+ return jnp.array(sigmas)
165
+
166
+ def _get_resume_params(self, params):
167
+ params['sigmas'] = self.sigmas_
168
+ return params
169
+
170
+ def _init_state(self, X, y, key):
171
+ key1, key2 = jax.random.split(key)
172
+ prototypes, proto_labels = self._init_prototypes(
173
+ X, y, self.n_prototypes_per_class, key1
174
+ )
175
+ sigmas = self._estimate_sigmas(X, y, prototypes, proto_labels)
176
+ params = {'prototypes': prototypes, 'sigmas': sigmas}
177
+ opt_state = self._optimizer.init(params)
178
+ from prosemble.models.prototype_base import SupervisedState
179
+ state = SupervisedState(
180
+ prototypes=prototypes,
181
+ opt_state=opt_state,
182
+ loss=jnp.array(float('inf')),
183
+ iteration=0,
184
+ converged=False,
185
+ )
186
+ return state, params, proto_labels
187
+
188
+ def _compute_loss(self, params, X, y, proto_labels):
189
+ prototypes = params['prototypes']
190
+ sigmas = jnp.maximum(params['sigmas'], self.sigma_min)
191
+ distances = kernel_distance_squared_per_proto(X, prototypes, sigmas)
192
+ from prosemble.core.losses import glvq_loss_with_transfer
193
+ return glvq_loss_with_transfer(
194
+ distances, y, proto_labels,
195
+ transfer_fn=self.transfer_fn,
196
+ margin=self.margin,
197
+ beta=self.beta,
198
+ )
199
+
200
+ def _post_update(self, params):
201
+ params['sigmas'] = jnp.maximum(params['sigmas'], self.sigma_min)
202
+ return params
203
+
204
+ def _extract_results(self, params, proto_labels, loss_history, n_iter, **kwargs):
205
+ super()._extract_results(params, proto_labels, loss_history, n_iter, **kwargs)
206
+ self.sigmas_ = params['sigmas']
207
+
208
+ @property
209
+ def kernel_bandwidths(self):
210
+ """Return the learned per-prototype bandwidths."""
211
+ if self.sigmas_ is None:
212
+ raise ValueError("Model not fitted. Call fit() first.")
213
+ return self.sigmas_
214
+
215
+ def predict(self, X):
216
+ """Predict using learned kernel distance."""
217
+ self._check_fitted()
218
+ X = jnp.asarray(X, dtype=jnp.float32)
219
+ return _predict_dkglvq_jit(
220
+ X, self.prototypes_, self.sigmas_,
221
+ self.prototype_labels_, self.sigma_min,
222
+ )
223
+
224
+ def _get_quantizable_attrs(self):
225
+ attrs = super()._get_quantizable_attrs()
226
+ if self.sigmas_ is not None:
227
+ attrs.append('sigmas_')
228
+ return attrs
229
+
230
+ def _get_fitted_arrays(self):
231
+ arrays = super()._get_fitted_arrays()
232
+ if self.sigmas_ is not None:
233
+ arrays['sigmas_'] = np.asarray(self.sigmas_)
234
+ return arrays
235
+
236
+ def _set_fitted_arrays(self, arrays):
237
+ super()._set_fitted_arrays(arrays)
238
+ if 'sigmas_' in arrays:
239
+ self.sigmas_ = jnp.asarray(arrays['sigmas_'])
240
+
241
+ def _get_hyperparams(self):
242
+ hp = super()._get_hyperparams()
243
+ hp['beta'] = self.beta
244
+ hp['sigma_init'] = self.sigma_init
245
+ hp['sigma_min'] = self.sigma_min
246
+ return hp