GMM implementation in Python
3 open threads
3 open threads
Implementation of GMM machine and trainer in Python using dask arrays
Edited by Yannick DAYER
Merge request reports
Activity
added 5 commits
Toggle commit listadded 15 commits
- afa11cd2 - Start of port of K-Means to python
- 5e6e3289 - k-means implementation and tests
- 02a68241 - [black]
- 91755436 - Finalize and optimize k-means
- c5126147 - Moved kmeans to cluster folder
- ee4a2ba1 - KMeansMachine as Transformer
- e40c0ea7 - Tests for python k-means
- 510e8517 - Plot script to use python k-means
- 4c21f381 - Update the cluster module init
- 929d13f7 - First draw for GMM machine and trainers
- 2e96065d - Implement GMM log_likelihood computation in python
- 20dcd079 - Sum the likelihood along the gaussian means first
- 67ae29ba - Fix the picklability test for GMM
- 2144b866 - Finalize GMM Machine log likelihood computation
- b4b0a14d - GMMMachine as Transformer
Toggle commit listadded 2 commits
added 2 commits
Hi @ydayer could you please rebase this?
added 25 commits
-
705f8fa8...8119eed1 - 6 commits from branch
pure-python - 63b46d69 - First draw for GMM machine and trainers
- 91554e3e - Implement GMM log_likelihood computation in python
- 9c308c5c - Sum the likelihood along the gaussian means first
- e523aae3 - Fix the picklability test for GMM
- 26f3a1eb - Finalize GMM Machine log likelihood computation
- 83d03586 - GMMMachine as Transformer
- 7785edfd - Gaussian and Statistics classes
- f4322d1f - GMM e step progress
- 78b00908 - e_step continued and tests [wip]
- 3a30fcea - e_step done
- 18de1129 - Fix Gaussian mean naming, use structuredarrays
- a96b19d5 - Tweaks to weights calculation
- cac68841 - GMM stats now has an add operator
- 580c1ec0 - Adding the ML GMM m_step and machine's log_weights
- 8d0b9a5d - Adding GMM tests
- 2fa45c05 - Fix Dask array not supporting str as index
- d788d72c - Reset to accept dask arrays
- 381222e8 - GMM port progress
- fe31aead - [doc] Plot code example of GMM
Toggle commit list-
705f8fa8...8119eed1 - 6 commits from branch
added 1 commit
- 07c854bc - Remove bob_devel from conda test requirements.
added 47 commits
-
07c854bc...6f4a29cb - 16 commits from branch
pure-python - 8b3ed094 - First draw for GMM machine and trainers
- 056ba52d - Implement GMM log_likelihood computation in python
- ca6181e7 - Sum the likelihood along the gaussian means first
- 46930972 - Fix the picklability test for GMM
- 7f6b97ad - Finalize GMM Machine log likelihood computation
- d0380069 - GMMMachine as Transformer
- 25a44748 - Gaussian and Statistics classes
- b8806f98 - GMM e step progress
- 9d518205 - e_step continued and tests [wip]
- a62390a5 - e_step done
- 910d67fe - Fix Gaussian mean naming, use structuredarrays
- a69ba178 - Tweaks to weights calculation
- b81c7682 - GMM stats now has an add operator
- 4a7e8f7f - Adding the ML GMM m_step and machine's log_weights
- 39e555ca - Adding GMM tests
- 1fff8c2f - Fix Dask array not supporting str as index
- 9d559cc0 - Reset to accept dask arrays
- e47b5733 - GMM port progress
- 53faf17f - [doc] Plot code example of GMM
- 62dc19f5 - Fix the GMM log_likelihood and add related tests.
- b081c11e - Remove the steps in GMM plot example
- 30e0bc81 - Added setters and getters for machine mean and var
- 2ed45f04 - Added MAP GMM implementation in python
- 50c6f25c - Added and adapted old GMM Machine tests
- 4938d2ab - Add gaussian tests back
- 87838908 - Variance threshold application at init for GMM
- a1942ad7 - Epsilon as default variance threshold
- 00bd4f45 - ML GMM test
- 6512af49 - MAP GMM fix and tests
- 356a3681 - Fixed test_gaussian not passing
- a530fdb5 - Remove Multi- and Gaussian, use Gaussians instead.
Toggle commit list-
07c854bc...6f4a29cb - 16 commits from branch
added 1 commit
- 4f3ca356 - Rename Statistics to GMMStats, tests for GMMStats.
added 2 commits
added 1 commit
- 219e6a46 - Start of `linear scoring` impl in `GMMMachine`.
added 1 commit
- 99e957f0 - Added linear_scoring; Added GMM Transformer tests.
added 1 commit
- 5f438c47 - Save and load in GMMMachine, linear_scoring as fct
added 1 commit
- 2ce9f1f1 - GMM gaussians init fixes, save and load fixes
added 1 commit
- e1f88a91 - Precomputing g_norms, tests fixes, h5py load/save.
added 1 commit
- c1c236c7 - [conda] copy meta.yaml from bob.devtools template.
assigned to @amohammadi and unassigned @ydayer
39 44 - numpy {{ numpy }} 40 45 - dask {{ dask }} 41 46 - dask-ml {{ dask_ml }} 47 - h5py {{ h5py }} 42 48 run: 43 49 - python 44 50 - setuptools 45 51 - boost 46 52 - {{ pin_compatible('numpy') }} 47 53 - {{ pin_compatible('dask') }} 48 54 - {{ pin_compatible('dask-ml') }} added 1 commit
- c07f0ce4 - Always keep state in numpy arrays instead of dask arrays
added 1 commit
- ebaae076 - Remove Gaussians class; use means, variances attr.
added 2 commits
added 1 commit
- d702d3d6 - Fix unsupported numpy array as `out` parameter.
added 11 commits
-
d702d3d6...092c914e - 9 commits from branch
pure-python - b4921b3d - Merge branch 'pure-python' into py-gmm
- 3d5a0df6 - persist the state of GMM between each training iteration
-
d702d3d6...092c914e - 9 commits from branch
added 2 commits
36 34 .. autosummary:: 37 35 38 36 bob.learn.em.cluster.KMeansMachine 37 bob.learn.em.mixture.GMMStats 38 bob.learn.em.mixture.GMMMachine 39 39 40 40 .. 41 41 TODO uncomment when implemented 42 bob.learn.em.Gaussian 43 bob.learn.em.GMMStats 44 bob.learn.em.GMMMachine mentioned in commit 32ee5e62
mentioned in issue bob.bio.base#174 (closed)
mentioned in issue bob.bio.gmm#27 (closed)
Please register or sign in to reply