polyergalio 0.1.0__tar.gz

polyergalio-0.1.0/CHANGES.md

@@ -0,0 +1,21 @@
+0.1.0
+=====================
+Renamed the package from ml_tools to polyergalio and prepared for PyPI
+publication:
+    - src/ml_tools became src/polyergalio, all imports updated
+    - pyproject.toml: dynamic version resolution, explicit src-layout
+      package discovery
+    - removed import-time logging
+
+0.0.1
+=====================
+Version algorithm aggregations - serialization and unserialize functions added
+    - Clusterings, SOM, PLSOM, free-SOM and CentNN modified
+    - Classification models updated with 
+    - NNet layers, network and DAG
+    - Generator "shapes patterns" added for clustering and images
+
+0.0.0 
+=====================
+Version zero - collecting algorithms and tools from my other repos and 
+projects.

polyergalio-0.1.0/LICENSE

@@ -0,0 +1,7 @@
+Copyright 2025
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

polyergalio-0.1.0/MANIFEST.in

@@ -0,0 +1,3 @@
+include LICENSE
+include README.md
+include CHANGES.md

polyergalio-0.1.0/PKG-INFO

@@ -0,0 +1,157 @@
+Metadata-Version: 2.4
+Name: polyergalio
+Version: 0.1.0
+Summary: From-scratch machine learning methods -- encoding, supervised learning, and clustering -- built on NumPy and SciPy
+Author-email: David Graey <graeyband@gmail.com>, Dr Charles Anderson <Chuck.Anderson@colostate.edu>
+License: MIT
+Project-URL: Homepage, https://github.com/Davidgraey/ml-tools
+Project-URL: Repository, https://github.com/Davidgraey/ml-tools
+Project-URL: Changelog, https://github.com/Davidgraey/ml-tools/blob/main/CHANGES.md
+Project-URL: Bug Tracker, https://github.com/Davidgraey/ml-tools/issues
+Keywords: machine-learning,numpy,clustering,self-organizing-map,neural-network
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: matplotlib
+Requires-Dist: numpy
+Requires-Dist: scipy
+Provides-Extra: gui
+Requires-Dist: matplotlib; extra == "gui"
+Provides-Extra: audio
+Requires-Dist: librosa; extra == "audio"
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Provides-Extra: dev
+Requires-Dist: pytest; extra == "dev"
+Requires-Dist: coverage; extra == "dev"
+Requires-Dist: build; extra == "dev"
+Requires-Dist: twine; extra == "dev"
+Dynamic: license-file
+
+# PolyerGalio
+![Python](https://img.shields.io/badge/python-3.10%2B-blue.svg)
+![NumPy](https://img.shields.io/badge/numpy-%3E%3D1.20-blue)
+![SciPy](https://img.shields.io/badge/scipy-%3E%3D1.7-blue)
+![License](https://img.shields.io/github/license/davidgraey/ml-tools)
+![Last Commit](https://img.shields.io/github/last-commit/davidgraey/ml-tools)
+![Repo Size](https://img.shields.io/github/repo-size/davidgraey/ml-tools)
+![Stars](https://img.shields.io/github/stars/davidgraey/ml-tools?style=social)
+
+---
+
+## Overview
+
+**PolyerGalio** (Greek for "multiple tools") is a collection of implemented machine learning methods ranging from  
+data encoding and processing pipelines to supervised learning and clustering.
+
+The focus of this repository is:
+- **Classic and alternative ML algorithms implemented with a unified 
+  interface**
+- **Numerical stability and performance**
+- **Novel extensions and original research contributions**
+
+All algorithms are implemented in **NumPy** and **SciPy**, with minimal external dependencies.
+
+---
+## Installation
+
+```bash
+pip install polyergalio
+```
+
+For local development:
+
+```bash
+pip install -e ".[test]"
+```
+
+---
+## Implemented Methods
+
+### 🔹 Encoding and Embedding Creation
+`src/polyergalio/encoders/*`
+- Categorical variable pipeline
+- Chronological variable (cyclical and absolute) pipeline
+- Numeric (normalized and raw) pipeline
+- Trainable Fourier Embedding pipeline
+- Trainable Text embedding pipeline
+- tokenization with sentencepiece
+
+### 🔹 Toy Dataset Generation
+`src/polyergalio/generators/*`
+
+### 🔹 Supervised Learning
+`src/polyergalio/models/supervised/*`
+#### Scaled Conjugate Gradient (SCG)
+- SCG for gradient descent applied to regression and logistic regression 
+  *(Møller); (Anderson)*
+- SCG regression with **Elastic Net regularization** *(novel)*
+- SCG classification:
+  - Binary
+  - Multinomial
+  - Multilabel
+
+#### Relative Weights (RW)
+- Johnson’s Relative Weights regression *(Johnson)*
+- Relative Weights applied to logistic regression  
+  *(Solís & Pasquier); (Tonidandel & LeBreton)*
+
+#### Tree Algorithms (EBM / EBTM)
+- Tree algorithms - Explainable Boosted-Tree Model (EBM)
+
+### 🔹 Unsupervised Learning & Clustering
+`src/polyergalio/models/clustering/*`
+#### Self-Organizing Maps
+- Self Organizing Maps, Parameterless Self-Organizing Maps - PLSOM  
+  *(Kohonen); (Berglund & Sitte)*
+  - Clustering and dimensionality reduction without hyperparameter adjustment
+- Growing Self Organizing Maps, Parameterless (grid)
+- Grid-free Growing Parameterless Self Organizing Maps
+  - FreeSOM - grows and shrinks under conditional updates
+  - *(novel)* fusion of Neural Gas and PLSOM
+
+
+#### Centroid Neural Networks (CENTNN)
+- Novel **Centroid Neural Network** for fast clustering and optimization  
+  *(Park, Dong-Chul)*
+- CENTNN with **N-dimensional density modeling**
+- *(novel)*
+
+---
+## Status
+
+**Active research / experimental**  
+APIs may change as methods are refined and extended.
+
+---
+## Authors and Contributors
+- "David Graey", "graeyband@gmail.com"
+- "Dr Charles Anderson", "Chuck.Anderson@colostate.edu"
+
+---
+## References
+
+Primary academic references are cited inline.  
+Full bibliographic references may be added in `/docs` in the future.
+
+https://packaging.python.org/en/latest/tutorials/packaging-projects/
+
+https://packaging.python.org/en/latest/tutorials/creating-documentation/
+
+
+# Visuals & Diagrams
+https://mermaid.js.org/config/Tutorials.html
+```mermaid
+flowchart LR;
+    A --> B;
+    A --> C;
+```

polyergalio-0.1.0/README.md

@@ -0,0 +1,118 @@
+# PolyerGalio
+![Python](https://img.shields.io/badge/python-3.10%2B-blue.svg)
+![NumPy](https://img.shields.io/badge/numpy-%3E%3D1.20-blue)
+![SciPy](https://img.shields.io/badge/scipy-%3E%3D1.7-blue)
+![License](https://img.shields.io/github/license/davidgraey/ml-tools)
+![Last Commit](https://img.shields.io/github/last-commit/davidgraey/ml-tools)
+![Repo Size](https://img.shields.io/github/repo-size/davidgraey/ml-tools)
+![Stars](https://img.shields.io/github/stars/davidgraey/ml-tools?style=social)
+
+---
+
+## Overview
+
+**PolyerGalio** (Greek for "multiple tools") is a collection of implemented machine learning methods ranging from  
+data encoding and processing pipelines to supervised learning and clustering.
+
+The focus of this repository is:
+- **Classic and alternative ML algorithms implemented with a unified 
+  interface**
+- **Numerical stability and performance**
+- **Novel extensions and original research contributions**
+
+All algorithms are implemented in **NumPy** and **SciPy**, with minimal external dependencies.
+
+---
+## Installation
+
+```bash
+pip install polyergalio
+```
+
+For local development:
+
+```bash
+pip install -e ".[test]"
+```
+
+---
+## Implemented Methods
+
+### 🔹 Encoding and Embedding Creation
+`src/polyergalio/encoders/*`
+- Categorical variable pipeline
+- Chronological variable (cyclical and absolute) pipeline
+- Numeric (normalized and raw) pipeline
+- Trainable Fourier Embedding pipeline
+- Trainable Text embedding pipeline
+- tokenization with sentencepiece
+
+### 🔹 Toy Dataset Generation
+`src/polyergalio/generators/*`
+
+### 🔹 Supervised Learning
+`src/polyergalio/models/supervised/*`
+#### Scaled Conjugate Gradient (SCG)
+- SCG for gradient descent applied to regression and logistic regression 
+  *(Møller); (Anderson)*
+- SCG regression with **Elastic Net regularization** *(novel)*
+- SCG classification:
+  - Binary
+  - Multinomial
+  - Multilabel
+
+#### Relative Weights (RW)
+- Johnson’s Relative Weights regression *(Johnson)*
+- Relative Weights applied to logistic regression  
+  *(Solís & Pasquier); (Tonidandel & LeBreton)*
+
+#### Tree Algorithms (EBM / EBTM)
+- Tree algorithms - Explainable Boosted-Tree Model (EBM)
+
+### 🔹 Unsupervised Learning & Clustering
+`src/polyergalio/models/clustering/*`
+#### Self-Organizing Maps
+- Self Organizing Maps, Parameterless Self-Organizing Maps - PLSOM  
+  *(Kohonen); (Berglund & Sitte)*
+  - Clustering and dimensionality reduction without hyperparameter adjustment
+- Growing Self Organizing Maps, Parameterless (grid)
+- Grid-free Growing Parameterless Self Organizing Maps
+  - FreeSOM - grows and shrinks under conditional updates
+  - *(novel)* fusion of Neural Gas and PLSOM
+
+
+#### Centroid Neural Networks (CENTNN)
+- Novel **Centroid Neural Network** for fast clustering and optimization  
+  *(Park, Dong-Chul)*
+- CENTNN with **N-dimensional density modeling**
+- *(novel)*
+
+---
+## Status
+
+**Active research / experimental**  
+APIs may change as methods are refined and extended.
+
+---
+## Authors and Contributors
+- "David Graey", "graeyband@gmail.com"
+- "Dr Charles Anderson", "Chuck.Anderson@colostate.edu"
+
+---
+## References
+
+Primary academic references are cited inline.  
+Full bibliographic references may be added in `/docs` in the future.
+
+https://packaging.python.org/en/latest/tutorials/packaging-projects/
+
+https://packaging.python.org/en/latest/tutorials/creating-documentation/
+
+
+# Visuals & Diagrams
+https://mermaid.js.org/config/Tutorials.html
+```mermaid
+flowchart LR;
+    A --> B;
+    A --> C;
+```

polyergalio-0.1.0/pyproject.toml

@@ -0,0 +1,69 @@
+[build-system]
+requires = ["setuptools >= 61.0"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "polyergalio"
+dynamic = ["version"]
+authors = [
+    {name = "David Graey", email = "graeyband@gmail.com"},
+    {name = "Dr Charles Anderson", email = "Chuck.Anderson@colostate.edu"}
+]
+dependencies = [
+    "matplotlib",
+    "numpy",
+    "scipy"
+]
+
+description = "From-scratch machine learning methods -- encoding, supervised learning, and clustering -- built on NumPy and SciPy"
+readme = "README.md"
+requires-python = ">=3.10"
+license = {text = "MIT"}
+keywords = ["machine-learning", "numpy", "clustering", "self-organizing-map", "neural-network"]
+
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Science/Research",
+    "License :: OSI Approved :: MIT License",
+    "Operating System :: OS Independent",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.10",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Topic :: Scientific/Engineering :: Artificial Intelligence",
+]
+
+[project.optional-dependencies]
+gui = ["matplotlib"]
+audio = ["librosa"]
+test = ["pytest"]
+dev = ["pytest", "coverage", "build", "twine"]
+
+[tool.setuptools.dynamic]
+version = {attr = "polyergalio.__version__"}
+
+[tool.setuptools.packages.find]
+where = ["src"]
+
+[tool.coverage.run]
+source = ["src/polyergalio"]
+branch = true
+
+[tool.coverage.report]
+show_missing = true
+skip_covered = true
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+# src layout, so tests import polyergalio without an editable install
+pythonpath = ["src", "tests"]
+addopts = "--strict-markers"
+markers = [
+    "slow: finite-difference gradient checks, run with -m 'not slow' to skip",
+]
+
+[project.urls]
+Homepage = "https://github.com/Davidgraey/ml-tools"
+Repository = "https://github.com/Davidgraey/ml-tools"
+Changelog = "https://github.com/Davidgraey/ml-tools/blob/main/CHANGES.md"
+"Bug Tracker" = "https://github.com/Davidgraey/ml-tools/issues"

polyergalio-0.1.0/setup.cfg

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

polyergalio-0.1.0/src/polyergalio/__init__.py

@@ -0,0 +1,12 @@
+"""
+polyergalio: implemented machine learning methods -- encoding,
+supervised learning, and clustering -- built on NumPy and SciPy.
+"""
+import logging
+
+__version__ = "0.1.0"
+
+log = logging.getLogger(__name__)
+log.addHandler(logging.NullHandler())
+
+__all__ = ["__version__"]

polyergalio-0.1.0/src/polyergalio/distances.py

@@ -0,0 +1,112 @@
+import numpy as np
+from numpy.typing import NDArray
+from scipy.spatial.distance import cdist, pdist
+from polyergalio.utilities import preformat_expected_shapes
+from polyergalio.models.constants import EPSILON
+
+
+########### DISTANCES ###########
+def manhattan_distance(x: NDArray, y: NDArray, summed: bool = True) -> NDArray:
+    """
+    aka Chebyshev
+    :param x:
+    :param y:
+    :return:
+    """
+    x, y = preformat_expected_shapes(x, y)
+    # sum(np.abs(x-y), axis=?)
+    if summed:
+        return np.sum(np.abs(x - y), axis=-1)
+    else:
+        return np.abs(x - y)
+
+
+def grid_manhattan_distance(row_a, col_a, row_b, col_b):
+    return abs(row_a - row_b) + abs(col_a - col_b)
+
+
+def euclidian_distance(x: NDArray, y: NDArray) -> NDArray:
+    """
+    for size (5000, 100), per sample, np.sqrtsum method took 1.07ms
+    linalg.norm took 1.21ms
+    numpy 1.20.3
+    :param x:
+    :param y:
+    :return:
+    """
+    x, y = preformat_expected_shapes(x, y)
+    return np.sqrt(np.sum((x - y) ** 2, axis=-1))  # + EPSILON
+    # return cdist(x, y, metric='euclidean')
+
+
+def norm_euclidian_distance(x: NDArray, y: NDArray) -> NDArray:
+    # numerator = (np.linalg.norm((x - np.mean(x)) - (y - np.mean(y))) ** 2)
+    # denom = (np.linalg.norm(x - np.mean(x)) ** 2 + np.linalg.norm(y - np.mean(y)) ** 2)
+    # return 0.5 * (numerator / denom)
+
+    vx = np.var(x, axis=-1)
+    vy = np.var(y, axis=-1)
+    return 0.5 * (vx / (vy + vx))
+
+
+def mahalonobis_distance(x: NDArray, y: NDArray) -> NDArray:
+    """
+
+    multivariate equivilant of euclidian distance, comparing point to distribution
+    :param x:
+    :param y:
+    :return:
+    """
+    stacked_vecs = np.vstack([x, y])
+    covariance = np.cov(stacked_vecs.T)
+    inv_covariance = np.linalg.inv(covariance)
+    delta = x - y
+    # may have to fix reshape for multi-dim...
+    return np.sqrt(np.einsum('nj,jk,nk->n', delta, inv_covariance, delta)).reshape(x.shape[0], -1)
+
+
+def hamming_distance(x: NDArray, y: NDArray) -> NDArray:
+    """
+    :param x:
+    :param y:
+    :return:
+    """
+    # axis?
+    return np.count_nonzero(x != y, axis=-1)
+
+
+def cosine_distance(x: NDArray, y: NDArray) -> NDArray:
+    """
+    inverse cosine similarity
+    :param x:
+    :param y:
+    :return:
+    """
+    x, y = preformat_expected_shapes(x, y)
+    return 1 - cosine_similarity(x, y)
+
+
+# +---------Similarity---------
+def cosine_similarity(x: NDArray, y: NDArray) -> NDArray:
+    """
+    requires numpy v1.7+ for 'where' in np.divide()
+    numpy array is [samples, dimension_1, dimension_2, ...]
+    we will compute cosine distance per samples dimension
+        input array is [500, 5] - output will be [500, 1]
+    :param x:
+    :param y:
+    :return:
+    """
+    dot = np.sum(x * y, axis=-1)
+    x_norm = np.sqrt(np.sum(x * x, axis=-1))
+    y_norm = np.sqrt(np.sum(y * y, axis=-1))
+
+    denom = x_norm * y_norm
+    return dot / np.maximum(denom, EPSILON)
+
+
+def jaccard_similarity(x: NDArray, y: NDArray) -> NDArray:
+    """aka Jaccard Index
+    for use in boolean / on-hot / binary array
+    """
+    return np.bitwise_and(x, y).sum(axis=-1) / np.bitwise_or(x, y).sum(axis=-1)

polyergalio-0.1.0/src/polyergalio/encoders/__init__.py

@@ -0,0 +1,3 @@
+import logging
+
+log = logging.getLogger(__name__)