first implementation commit

[akilandrews]

First implementation commit for mini-batch loop optimization chassis. Succesfully ran Abseil unit test tests.optimize.LengthScaleTest ('lool' only) and tests.optimize.NuTest ('lool', 'mse', 'huber'):

... [ RUN ] LengthScaleTest.test_length_scale_mini_batch0 (250, 20, {'nn_method': 'exact', 'algorithm': 'ball_tree'}, ['lool', {}, 'analytic'], 'loo_crossval', ['bayesian', {'random_state': 1, 'init_points': 5, 'n_iter': 20, 'allow_duplicate_points': True}]) optimizes length_scale with median relative squared error 0.06781370402848766 [ OK ] LengthScaleTest.test_length_scale_mini_batch0 (250, 20, {'nn_method': 'exact', 'algorithm': 'ball_tree'}, ['lool', {}, 'analytic'], 'loo_crossval', ['bayesian', {'random_state': 1, 'init_points': 5, 'n_iter': 20, 'allow_duplicate_points': True}]) ... [ RUN ] NuTest.test_nu_mini_batch0 (250, 20, {'nn_method': 'exact', 'algorithm': 'ball_tree'}, ['lool', {}, 'analytic'], 'loo_crossval', ['bayesian', {'random_state': 1, 'init_points': 3, 'n_iter': 10, 'allow_duplicate_points': True}]) optimizes nu with mean relative squared error 0.004897727480108819 [ OK ] NuTest.test_nu_mini_batch0 (250, 20, {'nn_method': 'exact', 'algorithm': 'ball_tree'}, ['lool', {}, 'analytic'], 'loo_crossval', ['bayesian', {'random_state': 1, 'init_points': 3, 'n_iter': 10, 'allow_duplicate_points': True}]) [ RUN ] NuTest.test_nu_mini_batch1 (250, 20, {'nn_method': 'exact', 'algorithm': 'ball_tree'}, ['mse', {}, None], 'loo_crossval', ['bayesian', {'random_state': 1, 'init_points': 3, 'n_iter': 10, 'allow_duplicate_points': True}]) Data point [1.] is not unique. 1 duplicates registered. Continuing ... optimizes nu with mean relative squared error 0.012607040388286268 [ OK ] NuTest.test_nu_mini_batch1 (250, 20, {'nn_method': 'exact', 'algorithm': 'ball_tree'}, ['mse', {}, None], 'loo_crossval', ['bayesian', {'random_state': 1, 'init_points': 3, 'n_iter': 10, 'allow_duplicate_points': True}]) [ RUN ] NuTest.test_nu_mini_batch2 (250, 20, {'nn_method': 'exact', 'algorithm': 'ball_tree'}, ['huber', {'boundary_scale': 1.5}, None], 'loo_crossval', ['bayesian', {'random_state': 1, 'init_points': 3, 'n_iter': 10, 'allow_duplicate_points': True}]) optimizes nu with mean relative squared error 0.02018999864653846 [ OK ] NuTest.test_nu_mini_batch2 (250, 20, {'nn_method': 'exact', 'algorithm': 'ball_tree'}, ['huber', {'boundary_scale': 1.5}, None], 'loo_crossval', ['bayesian', {'random_state': 1, 'init_points': 3, 'n_iter': 10, 'allow_duplicate_points': True}])

Ran 12 tests in 847.545s

OK