BayesHalvingSearchCV 0.1.0__tar.gz

bayeshalvingsearchcv-0.1.0/LICENSE

@@ -0,0 +1,29 @@
+BSD 3-Clause License
+
+Copyright (c) 2026, Rodrigo Pinto Coelho
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this
+   list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice,
+   this list of conditions and the following disclaimer in the documentation
+   and/or other materials provided with the distribution.
+
+3. Neither the name of the copyright holder nor the names of its
+   contributors may be used to endorse or promote products derived from
+   this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

bayeshalvingsearchcv-0.1.0/PKG-INFO

@@ -0,0 +1,141 @@
+Metadata-Version: 2.4
+Name: BayesHalvingSearchCV
+Version: 0.1.0
+Summary: scikit-learn hyperparameter search with multi-fidelity data growth and scatter-search multi-start: a from-scratch Bayesian (GP + Expected Improvement) search, BayesHalvingSearchCV, plus a Hooke-Jeeves pattern search, PatternSearchCV
+Author: Rodrigo Pinto Coelho
+License-Expression: BSD-3-Clause
+Project-URL: Homepage, https://github.com/rpcoelho17/PatternSearch_HyperParam_Opt__V2
+Project-URL: Repository, https://github.com/rpcoelho17/PatternSearch_HyperParam_Opt__V2
+Keywords: scikit-learn,hyperparameter-optimization,bayesian-optimization,pattern-search,multi-fidelity,gaussian-process,cross-validation,model-selection
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: Intended Audience :: Developers
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.24
+Requires-Dist: scipy>=1.10
+Requires-Dist: scikit-learn<1.10,>=1.6
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Provides-Extra: docs
+Requires-Dist: sphinx; extra == "docs"
+Requires-Dist: numpydoc; extra == "docs"
+Requires-Dist: pydata-sphinx-theme; extra == "docs"
+Dynamic: license-file
+
+# BayesHalvingSearchCV & PatternSearchCV
+
+During my post-graduate studies in Data Science at USI Switzerland, I had to hyperparameter-tune
+models on genuinely large datasets, and I went looking for a way to find good
+hyperparameters quickly instead of paying full price for every fit.
+Inspired by MATLAB's Global Optimization Toolbox pattern search for discrete
+problems, I built a first version of what became `PatternSearchCV` — and even
+that first version already beat Bayesian search on real benchmarks, simply
+because it deduplicated repeated evaluations, something the Bayesian search
+tools I compared it against didn't do; that alone saved real fits on large
+datasets. I later added a "halving" mechanism on top of that base pattern
+search — growing how much data each evaluation sees as the search's own
+trajectory shows it converging (measured by the shrinking distance between
+successive improving moves), instead of committing to full data from the
+start — and that further improved the gains over the deduplication-only
+version. When I eventually benchmarked properly against Optuna, the picture
+became precise: Optuna's Bayesian search found the optimum in fewer raw
+trials, but every one of those trials evaluated 100% of the data — my
+advantage was coming from deduplication, the multi-fidelity data growth, and
+the halving logic together, not from having fewer trials. So I adapted that
+same multi-fidelity infrastructure onto a from-scratch Bayesian search of my
+own, giving `BayesHalvingSearchCV` Optuna-quality trial efficiency without
+paying full price for every trial.
+
+**Both estimators implement scikit-learn's standard search-CV interface and
+work with any scikit-learn-compatible estimator** — they drop into existing
+pipelines exactly the way `GridSearchCV` or `RandomizedSearchCV` do. Our own
+benchmarks use `ExtraTreesRegressor` specifically because it's fast and
+performs well on the dataset we tested against, not because either algorithm
+is limited to it.
+
+**Best measured results** (523K-row retail regression benchmark; full
+methodology and every run logged in [`EXPERIMENTS.md`](EXPERIMENTS.md)):
+`BayesHalvingSearchCV` reached the benchmark's historical optimum at
+**3.125 full-fit equivalents** (Experiment 17); `PatternSearchCV` reached the
+same optimum at **5.04 full-fit equivalents** (Experiment 11). For
+comparison, Optuna's own GP-based sampler needed **15.00 full-fit
+equivalents** — every one of its trials at 100% of the data — to reach a
+comparable answer. That's a **4.8× reduction** in compute for
+`BayesHalvingSearchCV` and a **2.98× reduction** for `PatternSearchCV`,
+relative to Optuna, for the same answer.
+
+GPU acceleration is available the same way it is for any scikit-learn
+estimator: enable it in `estimator`'s own settings. See
+[`API_REFERENCE.md`](API_REFERENCE.md) for the full explanation.
+
+Both estimators share one **multi-fidelity "bullseye" data-growth mechanism**
+and one **scatter-search multi-start layer**:
+
+- `PatternSearchCV` — classic Hooke-Jeeves pattern search (1961), adapted to
+  grow its data budget as it converges.
+- `BayesHalvingSearchCV` — a from-scratch Gaussian Process + Expected
+  Improvement Bayesian search, on the exact same multi-fidelity
+  infrastructure, with **zero additional dependencies** (no Optuna, no
+  torch — just `numpy`, `scipy`, `scikit-learn`, already required by
+  `PatternSearchCV`).
+
+Both estimators start every search on a small, representative subsample of
+the training data and only pay full price once their own search trajectory
+shows them converging on an optimum — every reported result is still
+confirmed on 100% of the data before it's trusted.
+
+## Quick start
+
+```python
+from bayes_halving_search_cv import PatternSearchCV, BayesHalvingSearchCV
+from sklearn.model_selection import TimeSeriesSplit
+
+param_grid = {"max_depth": [3, 5, 7, 9, 12, 16], "min_samples_leaf": [1, 2, 4, 8]}
+
+search = PatternSearchCV(
+    estimator, param_grid,
+    cv=TimeSeriesSplit(n_splits=5),
+    scoring="neg_mean_absolute_error",
+    n_starts=4,               # scatter-search multi-start
+    subsample="stratified",   # transition sampling for time-series data
+    random_state=0,
+)
+search.fit(X, y)
+search.best_params_       # chosen ONLY from full-data evaluations
+search.local_optima_      # the map: every distinct optimum found
+search.cv_results_        # every point evaluated, and its score
+search.search_history_    # every confirmed-improving move across every start
+
+# or the Bayesian search, on the exact same multi-fidelity infrastructure:
+search = BayesHalvingSearchCV(estimator, param_grid, cv=TimeSeriesSplit(5),
+                              scoring="neg_mean_absolute_error", random_state=0)
+search.fit(X, y)
+```
+
+For the full parameter reference and worked examples — how to specify a
+search space, the data ladder, `contraction="eager"`'s cost/risk trade-off,
+and why `subsample="stratified"` matters for time series — see
+[`API_REFERENCE.md`](API_REFERENCE.md). For the full design rationale behind
+every default, see [`PatternSearchCV_SPEC.md`](PatternSearchCV_SPEC.md) and
+[`BAYESHALVINGSearchCV_SPEC.md`](BAYESHALVINGSearchCV_SPEC.md).
+
+## Development
+
+```
+python -m venv .venv
+.venv/Scripts/pip install -e .[test]
+.venv/Scripts/python -m pytest
+```
+
+Logging: the package logs every algorithmic decision (moves, contractions,
+ring calibrations and crossings, data climbs, merges, cache statistics) to the
+`SearchCV` logger. `verbose=1` attaches a stream handler at INFO,
+`verbose=2` at DEBUG.

bayeshalvingsearchcv-0.1.0/README.md

@@ -0,0 +1,109 @@
+# BayesHalvingSearchCV & PatternSearchCV
+
+During my post-graduate studies in Data Science at USI Switzerland, I had to hyperparameter-tune
+models on genuinely large datasets, and I went looking for a way to find good
+hyperparameters quickly instead of paying full price for every fit.
+Inspired by MATLAB's Global Optimization Toolbox pattern search for discrete
+problems, I built a first version of what became `PatternSearchCV` — and even
+that first version already beat Bayesian search on real benchmarks, simply
+because it deduplicated repeated evaluations, something the Bayesian search
+tools I compared it against didn't do; that alone saved real fits on large
+datasets. I later added a "halving" mechanism on top of that base pattern
+search — growing how much data each evaluation sees as the search's own
+trajectory shows it converging (measured by the shrinking distance between
+successive improving moves), instead of committing to full data from the
+start — and that further improved the gains over the deduplication-only
+version. When I eventually benchmarked properly against Optuna, the picture
+became precise: Optuna's Bayesian search found the optimum in fewer raw
+trials, but every one of those trials evaluated 100% of the data — my
+advantage was coming from deduplication, the multi-fidelity data growth, and
+the halving logic together, not from having fewer trials. So I adapted that
+same multi-fidelity infrastructure onto a from-scratch Bayesian search of my
+own, giving `BayesHalvingSearchCV` Optuna-quality trial efficiency without
+paying full price for every trial.
+
+**Both estimators implement scikit-learn's standard search-CV interface and
+work with any scikit-learn-compatible estimator** — they drop into existing
+pipelines exactly the way `GridSearchCV` or `RandomizedSearchCV` do. Our own
+benchmarks use `ExtraTreesRegressor` specifically because it's fast and
+performs well on the dataset we tested against, not because either algorithm
+is limited to it.
+
+**Best measured results** (523K-row retail regression benchmark; full
+methodology and every run logged in [`EXPERIMENTS.md`](EXPERIMENTS.md)):
+`BayesHalvingSearchCV` reached the benchmark's historical optimum at
+**3.125 full-fit equivalents** (Experiment 17); `PatternSearchCV` reached the
+same optimum at **5.04 full-fit equivalents** (Experiment 11). For
+comparison, Optuna's own GP-based sampler needed **15.00 full-fit
+equivalents** — every one of its trials at 100% of the data — to reach a
+comparable answer. That's a **4.8× reduction** in compute for
+`BayesHalvingSearchCV` and a **2.98× reduction** for `PatternSearchCV`,
+relative to Optuna, for the same answer.
+
+GPU acceleration is available the same way it is for any scikit-learn
+estimator: enable it in `estimator`'s own settings. See
+[`API_REFERENCE.md`](API_REFERENCE.md) for the full explanation.
+
+Both estimators share one **multi-fidelity "bullseye" data-growth mechanism**
+and one **scatter-search multi-start layer**:
+
+- `PatternSearchCV` — classic Hooke-Jeeves pattern search (1961), adapted to
+  grow its data budget as it converges.
+- `BayesHalvingSearchCV` — a from-scratch Gaussian Process + Expected
+  Improvement Bayesian search, on the exact same multi-fidelity
+  infrastructure, with **zero additional dependencies** (no Optuna, no
+  torch — just `numpy`, `scipy`, `scikit-learn`, already required by
+  `PatternSearchCV`).
+
+Both estimators start every search on a small, representative subsample of
+the training data and only pay full price once their own search trajectory
+shows them converging on an optimum — every reported result is still
+confirmed on 100% of the data before it's trusted.
+
+## Quick start
+
+```python
+from bayes_halving_search_cv import PatternSearchCV, BayesHalvingSearchCV
+from sklearn.model_selection import TimeSeriesSplit
+
+param_grid = {"max_depth": [3, 5, 7, 9, 12, 16], "min_samples_leaf": [1, 2, 4, 8]}
+
+search = PatternSearchCV(
+    estimator, param_grid,
+    cv=TimeSeriesSplit(n_splits=5),
+    scoring="neg_mean_absolute_error",
+    n_starts=4,               # scatter-search multi-start
+    subsample="stratified",   # transition sampling for time-series data
+    random_state=0,
+)
+search.fit(X, y)
+search.best_params_       # chosen ONLY from full-data evaluations
+search.local_optima_      # the map: every distinct optimum found
+search.cv_results_        # every point evaluated, and its score
+search.search_history_    # every confirmed-improving move across every start
+
+# or the Bayesian search, on the exact same multi-fidelity infrastructure:
+search = BayesHalvingSearchCV(estimator, param_grid, cv=TimeSeriesSplit(5),
+                              scoring="neg_mean_absolute_error", random_state=0)
+search.fit(X, y)
+```
+
+For the full parameter reference and worked examples — how to specify a
+search space, the data ladder, `contraction="eager"`'s cost/risk trade-off,
+and why `subsample="stratified"` matters for time series — see
+[`API_REFERENCE.md`](API_REFERENCE.md). For the full design rationale behind
+every default, see [`PatternSearchCV_SPEC.md`](PatternSearchCV_SPEC.md) and
+[`BAYESHALVINGSearchCV_SPEC.md`](BAYESHALVINGSearchCV_SPEC.md).
+
+## Development
+
+```
+python -m venv .venv
+.venv/Scripts/pip install -e .[test]
+.venv/Scripts/python -m pytest
+```
+
+Logging: the package logs every algorithmic decision (moves, contractions,
+ring calibrations and crossings, data climbs, merges, cache statistics) to the
+`SearchCV` logger. `verbose=1` attaches a stream handler at INFO,
+`verbose=2` at DEBUG.

bayeshalvingsearchcv-0.1.0/pyproject.toml

@@ -0,0 +1,53 @@
+[build-system]
+requires = ["setuptools>=77"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "BayesHalvingSearchCV"
+version = "0.1.0"
+description = "scikit-learn hyperparameter search with multi-fidelity data growth and scatter-search multi-start: a from-scratch Bayesian (GP + Expected Improvement) search, BayesHalvingSearchCV, plus a Hooke-Jeeves pattern search, PatternSearchCV"
+readme = "README.md"
+requires-python = ">=3.10"
+license = "BSD-3-Clause"
+license-files = ["LICENSE"]
+authors = [{ name = "Rodrigo Pinto Coelho" }]
+keywords = [
+    "scikit-learn",
+    "hyperparameter-optimization",
+    "bayesian-optimization",
+    "pattern-search",
+    "multi-fidelity",
+    "gaussian-process",
+    "cross-validation",
+    "model-selection",
+]
+classifiers = [
+    "Development Status :: 4 - Beta",
+    "Intended Audience :: Science/Research",
+    "Intended Audience :: Developers",
+    "Topic :: Scientific/Engineering :: Artificial Intelligence",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.10",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Operating System :: OS Independent",
+]
+dependencies = [
+    "numpy>=1.24",
+    "scipy>=1.10",
+    "scikit-learn>=1.6,<1.10",
+]
+
+[project.urls]
+Homepage = "https://github.com/rpcoelho17/PatternSearch_HyperParam_Opt__V2"
+Repository = "https://github.com/rpcoelho17/PatternSearch_HyperParam_Opt__V2"
+
+[project.optional-dependencies]
+test = ["pytest"]
+docs = ["sphinx", "numpydoc", "pydata-sphinx-theme"]
+
+[tool.setuptools.packages.find]
+where = ["src"]
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]

bayeshalvingsearchcv-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

bayeshalvingsearchcv-0.1.0/src/BayesHalvingSearchCV.egg-info/PKG-INFO

@@ -0,0 +1,141 @@
+Metadata-Version: 2.4
+Name: BayesHalvingSearchCV
+Version: 0.1.0
+Summary: scikit-learn hyperparameter search with multi-fidelity data growth and scatter-search multi-start: a from-scratch Bayesian (GP + Expected Improvement) search, BayesHalvingSearchCV, plus a Hooke-Jeeves pattern search, PatternSearchCV
+Author: Rodrigo Pinto Coelho
+License-Expression: BSD-3-Clause
+Project-URL: Homepage, https://github.com/rpcoelho17/PatternSearch_HyperParam_Opt__V2
+Project-URL: Repository, https://github.com/rpcoelho17/PatternSearch_HyperParam_Opt__V2
+Keywords: scikit-learn,hyperparameter-optimization,bayesian-optimization,pattern-search,multi-fidelity,gaussian-process,cross-validation,model-selection
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: Intended Audience :: Developers
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy>=1.24
+Requires-Dist: scipy>=1.10
+Requires-Dist: scikit-learn<1.10,>=1.6
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Provides-Extra: docs
+Requires-Dist: sphinx; extra == "docs"
+Requires-Dist: numpydoc; extra == "docs"
+Requires-Dist: pydata-sphinx-theme; extra == "docs"
+Dynamic: license-file
+
+# BayesHalvingSearchCV & PatternSearchCV
+
+During my post-graduate studies in Data Science at USI Switzerland, I had to hyperparameter-tune
+models on genuinely large datasets, and I went looking for a way to find good
+hyperparameters quickly instead of paying full price for every fit.
+Inspired by MATLAB's Global Optimization Toolbox pattern search for discrete
+problems, I built a first version of what became `PatternSearchCV` — and even
+that first version already beat Bayesian search on real benchmarks, simply
+because it deduplicated repeated evaluations, something the Bayesian search
+tools I compared it against didn't do; that alone saved real fits on large
+datasets. I later added a "halving" mechanism on top of that base pattern
+search — growing how much data each evaluation sees as the search's own
+trajectory shows it converging (measured by the shrinking distance between
+successive improving moves), instead of committing to full data from the
+start — and that further improved the gains over the deduplication-only
+version. When I eventually benchmarked properly against Optuna, the picture
+became precise: Optuna's Bayesian search found the optimum in fewer raw
+trials, but every one of those trials evaluated 100% of the data — my
+advantage was coming from deduplication, the multi-fidelity data growth, and
+the halving logic together, not from having fewer trials. So I adapted that
+same multi-fidelity infrastructure onto a from-scratch Bayesian search of my
+own, giving `BayesHalvingSearchCV` Optuna-quality trial efficiency without
+paying full price for every trial.
+
+**Both estimators implement scikit-learn's standard search-CV interface and
+work with any scikit-learn-compatible estimator** — they drop into existing
+pipelines exactly the way `GridSearchCV` or `RandomizedSearchCV` do. Our own
+benchmarks use `ExtraTreesRegressor` specifically because it's fast and
+performs well on the dataset we tested against, not because either algorithm
+is limited to it.
+
+**Best measured results** (523K-row retail regression benchmark; full
+methodology and every run logged in [`EXPERIMENTS.md`](EXPERIMENTS.md)):
+`BayesHalvingSearchCV` reached the benchmark's historical optimum at
+**3.125 full-fit equivalents** (Experiment 17); `PatternSearchCV` reached the
+same optimum at **5.04 full-fit equivalents** (Experiment 11). For
+comparison, Optuna's own GP-based sampler needed **15.00 full-fit
+equivalents** — every one of its trials at 100% of the data — to reach a
+comparable answer. That's a **4.8× reduction** in compute for
+`BayesHalvingSearchCV` and a **2.98× reduction** for `PatternSearchCV`,
+relative to Optuna, for the same answer.
+
+GPU acceleration is available the same way it is for any scikit-learn
+estimator: enable it in `estimator`'s own settings. See
+[`API_REFERENCE.md`](API_REFERENCE.md) for the full explanation.
+
+Both estimators share one **multi-fidelity "bullseye" data-growth mechanism**
+and one **scatter-search multi-start layer**:
+
+- `PatternSearchCV` — classic Hooke-Jeeves pattern search (1961), adapted to
+  grow its data budget as it converges.
+- `BayesHalvingSearchCV` — a from-scratch Gaussian Process + Expected
+  Improvement Bayesian search, on the exact same multi-fidelity
+  infrastructure, with **zero additional dependencies** (no Optuna, no
+  torch — just `numpy`, `scipy`, `scikit-learn`, already required by
+  `PatternSearchCV`).
+
+Both estimators start every search on a small, representative subsample of
+the training data and only pay full price once their own search trajectory
+shows them converging on an optimum — every reported result is still
+confirmed on 100% of the data before it's trusted.
+
+## Quick start
+
+```python
+from bayes_halving_search_cv import PatternSearchCV, BayesHalvingSearchCV
+from sklearn.model_selection import TimeSeriesSplit
+
+param_grid = {"max_depth": [3, 5, 7, 9, 12, 16], "min_samples_leaf": [1, 2, 4, 8]}
+
+search = PatternSearchCV(
+    estimator, param_grid,
+    cv=TimeSeriesSplit(n_splits=5),
+    scoring="neg_mean_absolute_error",
+    n_starts=4,               # scatter-search multi-start
+    subsample="stratified",   # transition sampling for time-series data
+    random_state=0,
+)
+search.fit(X, y)
+search.best_params_       # chosen ONLY from full-data evaluations
+search.local_optima_      # the map: every distinct optimum found
+search.cv_results_        # every point evaluated, and its score
+search.search_history_    # every confirmed-improving move across every start
+
+# or the Bayesian search, on the exact same multi-fidelity infrastructure:
+search = BayesHalvingSearchCV(estimator, param_grid, cv=TimeSeriesSplit(5),
+                              scoring="neg_mean_absolute_error", random_state=0)
+search.fit(X, y)
+```
+
+For the full parameter reference and worked examples — how to specify a
+search space, the data ladder, `contraction="eager"`'s cost/risk trade-off,
+and why `subsample="stratified"` matters for time series — see
+[`API_REFERENCE.md`](API_REFERENCE.md). For the full design rationale behind
+every default, see [`PatternSearchCV_SPEC.md`](PatternSearchCV_SPEC.md) and
+[`BAYESHALVINGSearchCV_SPEC.md`](BAYESHALVINGSearchCV_SPEC.md).
+
+## Development
+
+```
+python -m venv .venv
+.venv/Scripts/pip install -e .[test]
+.venv/Scripts/python -m pytest
+```
+
+Logging: the package logs every algorithmic decision (moves, contractions,
+ring calibrations and crossings, data climbs, merges, cache statistics) to the
+`SearchCV` logger. `verbose=1` attaches a stream handler at INFO,
+`verbose=2` at DEBUG.

bayeshalvingsearchcv-0.1.0/src/BayesHalvingSearchCV.egg-info/SOURCES.txt

@@ -0,0 +1,26 @@
+LICENSE
+README.md
+pyproject.toml
+src/BayesHalvingSearchCV.egg-info/PKG-INFO
+src/BayesHalvingSearchCV.egg-info/SOURCES.txt
+src/BayesHalvingSearchCV.egg-info/dependency_links.txt
+src/BayesHalvingSearchCV.egg-info/requires.txt
+src/BayesHalvingSearchCV.egg-info/top_level.txt
+src/bayes_halving_search_cv/__init__.py
+src/bayes_halving_search_cv/_bayes.py
+src/bayes_halving_search_cv/_climber.py
+src/bayes_halving_search_cv/_engine.py
+src/bayes_halving_search_cv/_fidelity.py
+src/bayes_halving_search_cv/_gp.py
+src/bayes_halving_search_cv/_sampling.py
+src/bayes_halving_search_cv/_search.py
+src/bayes_halving_search_cv/_space.py
+src/bayes_halving_search_cv/_starts.py
+tests/test_bayes.py
+tests/test_climber.py
+tests/test_estimator.py
+tests/test_fidelity.py
+tests/test_gp.py
+tests/test_sampling.py
+tests/test_sklearn_compat.py
+tests/test_starts.py

bayeshalvingsearchcv-0.1.0/src/BayesHalvingSearchCV.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

bayeshalvingsearchcv-0.1.0/src/BayesHalvingSearchCV.egg-info/requires.txt

@@ -0,0 +1,11 @@
+numpy>=1.24
+scipy>=1.10
+scikit-learn<1.10,>=1.6
+
+[docs]
+sphinx
+numpydoc
+pydata-sphinx-theme
+
+[test]
+pytest

bayeshalvingsearchcv-0.1.0/src/BayesHalvingSearchCV.egg-info/top_level.txt

@@ -0,0 +1 @@
+bayes_halving_search_cv

bayeshalvingsearchcv-0.1.0/src/bayes_halving_search_cv/__init__.py

@@ -0,0 +1,15 @@
+"""bayes-halving-search-cv: scikit-learn hyperparameter search with bullseye
+multi-fidelity data growth and scatter-search multi-start. Provides
+BayesHalvingSearchCV (a from-scratch GP + Expected Improvement Bayesian
+search) and PatternSearchCV (Hooke-Jeeves pattern search)."""
+
+import logging
+
+from ._bayes import BayesHalvingSearchCV
+from ._search import PatternSearchCV
+from ._space import Dimension, Space
+
+__all__ = ["PatternSearchCV", "BayesHalvingSearchCV", "Space", "Dimension"]
+__version__ = "0.1.0"
+
+logging.getLogger("SearchCV").addHandler(logging.NullHandler())