<p align="center">
<!-- PyPI Version -->
<a href="https://pypi.org/project/ixai">
<img src="https://img.shields.io/pypi/v/ixai.svg?color=blue" alt="PyPi">
</a>
<!-- PyPI status -->
<a href="https://pypi.org/project/ixai">
<img src="https://img.shields.io/pypi/status/ixai.svg?color=blue" alt="PyPi_status
</a>
<!-- License -->
<a href="https://opensource.org/licenses/MIT">
<img src="https://img.shields.io/badge/License-MIT-brightgreen.svg" alt="mit_license">
</a>
</p>
# ixai: Incremental Explainable Artificial Intelligence
This is the first iteration of our incremental explanation package.
Currently, it includes two explanation methods: PFI and SAGE.
Please look at the examples in the `examples` directory.
Please help us in improving our work by contributing or pointing to issues. We will update this iteration soon with further information.
## 🛠Installation
**ixai** is intended to work with **Python 3.8 and above**. Installation can be done via `pip`:
```sh
pip install ixai
```
## 📊 Quickstart
### Basic Classification
```python
>>> from river.metrics import Accuracy
>>> from river.ensemble import AdaptiveRandomForestClassifier
>>> from river.datasets.synth import Agrawal
>>> from ixai.explainer import IncrementalPFI
>>> stream = Agrawal(classification_function=2)
>>> feature_names = list([x_0 for x_0, _ in stream.take(1)][0].keys())
>>> model = AdaptiveRandomForestClassifier(n_models=10, max_depth=10, leaf_prediction='mc')
>>> incremental_pfi = IncrementalPFI(
... model_function=model.predict_one,
... loss_function=Accuracy(),
... feature_names=feature_names,
... smoothing_alpha=0.001,
... n_inner_samples=5
... )
>>> training_metric = Accuracy()
>>> for (n, (x, y)) in enumerate(stream, start=1)
... y_pred = model.predict_one(x) # inference
... training_metric.update(y, y_pred) # update score
... incremental_pfi.explain_one(x, y) # explaining
... model.learn_one(x, y) # learning
... if n % 1000 == 0:
... print(f"{n}: Accuracy: {training_metric.get():.3f}, PFI: {incremental_pfi.importance_values}")
1000: Accuracy: 0.785, PFI: {'age': 0.22, 'elevel': 0.18, 'zipcode': -0.07, 'salary': 0.04, 'commission': 0.05, 'loan': -0.06, 'car': 0.02, 'hyears': 0.03, 'hvalue': 0.03}
2000: Accuracy: 0.841, PFI: {'age': 0.26, 'elevel': 0.21, 'zipcode': -0.01, 'salary': 0.02, 'commission': 0.03, 'loan': -0.02, 'car': 0.02, 'hyears': 0.04, 'hvalue': 0.02}
3000: Accuracy: 0.921, PFI: {'age': 0.28, 'elevel': 0.24, 'zipcode': -0.00, 'salary': 0.00, 'commission': 0.01, 'loan': -0.01, 'car': 0.01, 'hyears': 0.01, 'hvalue': 0.00}
Raw data
{
"_id": null,
"home_page": "https://github.com/mmschlk/iXAI",
"name": "ixai",
"maintainer": "",
"docs_url": null,
"requires_python": ">=3.8.0",
"maintainer_email": "",
"keywords": "python,machine learning,online learning,xai",
"author": "Maximilian Muschalik",
"author_email": "maximilian.muschalik@ifi.lmu.de",
"download_url": "https://files.pythonhosted.org/packages/13/dd/d40dc8e088706df2dcfd209caf45919a5969e571cc75e3b3d6adacf53f6c/ixai-0.1.3.tar.gz",
"platform": null,
"description": "\n<p align=\"center\">\n <!-- PyPI Version -->\n <a href=\"https://pypi.org/project/ixai\">\n <img src=\"https://img.shields.io/pypi/v/ixai.svg?color=blue\" alt=\"PyPi\">\n </a>\n <!-- PyPI status -->\n <a href=\"https://pypi.org/project/ixai\">\n <img src=\"https://img.shields.io/pypi/status/ixai.svg?color=blue\" alt=\"PyPi_status\n </a>\n <!-- License -->\n <a href=\"https://opensource.org/licenses/MIT\">\n <img src=\"https://img.shields.io/badge/License-MIT-brightgreen.svg\" alt=\"mit_license\">\n </a>\n</p>\n\n# ixai: Incremental Explainable Artificial Intelligence\n\nThis is the first iteration of our incremental explanation package.\n\nCurrently, it includes two explanation methods: PFI and SAGE.\n\nPlease look at the examples in the `examples` directory.\n\nPlease help us in improving our work by contributing or pointing to issues. We will update this iteration soon with further information.\n\n## \ud83d\udee0 Installation\n**ixai** is intended to work with **Python 3.8 and above**. Installation can be done via `pip`:\n\n```sh\npip install ixai\n```\n\n## \ud83d\udcca Quickstart\n\n### Basic Classification\n```python\n>>> from river.metrics import Accuracy\n>>> from river.ensemble import AdaptiveRandomForestClassifier\n>>> from river.datasets.synth import Agrawal\n\n>>> from ixai.explainer import IncrementalPFI\n\n>>> stream = Agrawal(classification_function=2)\n>>> feature_names = list([x_0 for x_0, _ in stream.take(1)][0].keys())\n\n>>> model = AdaptiveRandomForestClassifier(n_models=10, max_depth=10, leaf_prediction='mc')\n\n>>> incremental_pfi = IncrementalPFI(\n... model_function=model.predict_one,\n... loss_function=Accuracy(),\n... feature_names=feature_names,\n... smoothing_alpha=0.001,\n... n_inner_samples=5\n... )\n\n>>> training_metric = Accuracy()\n>>> for (n, (x, y)) in enumerate(stream, start=1)\n... y_pred = model.predict_one(x) # inference\n... training_metric.update(y, y_pred) # update score\n... incremental_pfi.explain_one(x, y) # explaining\n... model.learn_one(x, y) # learning\n... if n % 1000 == 0:\n... print(f\"{n}: Accuracy: {training_metric.get():.3f}, PFI: {incremental_pfi.importance_values}\")\n\n1000: Accuracy: 0.785, PFI: {'age': 0.22, 'elevel': 0.18, 'zipcode': -0.07, 'salary': 0.04, 'commission': 0.05, 'loan': -0.06, 'car': 0.02, 'hyears': 0.03, 'hvalue': 0.03}\n2000: Accuracy: 0.841, PFI: {'age': 0.26, 'elevel': 0.21, 'zipcode': -0.01, 'salary': 0.02, 'commission': 0.03, 'loan': -0.02, 'car': 0.02, 'hyears': 0.04, 'hvalue': 0.02}\n3000: Accuracy: 0.921, PFI: {'age': 0.28, 'elevel': 0.24, 'zipcode': -0.00, 'salary': 0.00, 'commission': 0.01, 'loan': -0.01, 'car': 0.01, 'hyears': 0.01, 'hvalue': 0.00}\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Explainable Artificial Intelligence for dynamic and Incremental models.",
"version": "0.1.3",
"split_keywords": [
"python",
"machine learning",
"online learning",
"xai"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "477cd44a59ff2c423e25fd797e3ee48cb60926b7dc927c489192ccecaaf2ee88",
"md5": "04b65d080c9385b5dc180ec8ee2dbe6e",
"sha256": "fa9fb4cda178d642ad5929d36a45752cef367150c64128ebf8c43adf9619eed7"
},
"downloads": -1,
"filename": "ixai-0.1.3-py3-none-any.whl",
"has_sig": false,
"md5_digest": "04b65d080c9385b5dc180ec8ee2dbe6e",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.8.0",
"size": 46757,
"upload_time": "2023-01-18T15:23:54",
"upload_time_iso_8601": "2023-01-18T15:23:54.368531Z",
"url": "https://files.pythonhosted.org/packages/47/7c/d44a59ff2c423e25fd797e3ee48cb60926b7dc927c489192ccecaaf2ee88/ixai-0.1.3-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "13ddd40dc8e088706df2dcfd209caf45919a5969e571cc75e3b3d6adacf53f6c",
"md5": "e3db08ba18d50d48190e4a457f52b420",
"sha256": "a91b62bb106c66aa01829665189e28dd78242c00102048c3c5c55a58959c3902"
},
"downloads": -1,
"filename": "ixai-0.1.3.tar.gz",
"has_sig": false,
"md5_digest": "e3db08ba18d50d48190e4a457f52b420",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8.0",
"size": 30292,
"upload_time": "2023-01-18T15:23:56",
"upload_time_iso_8601": "2023-01-18T15:23:56.059818Z",
"url": "https://files.pythonhosted.org/packages/13/dd/d40dc8e088706df2dcfd209caf45919a5969e571cc75e3b3d6adacf53f6c/ixai-0.1.3.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-01-18 15:23:56",
"github": true,
"gitlab": false,
"bitbucket": false,
"github_user": "mmschlk",
"github_project": "iXAI",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "ixai"
}
JSON
