The Howso Engine™ is a natively and fully explainable ML engine, serving as an alternative to black box AI neural networks. Its core functionality gives users data exploration and machine learning capabilities through the creation and use of Trainees that help users store, explore, and analyze the relationships in their data, as well as make understandable, debuggable predictions. Howso leverages an instance-based learning approach with strong ties to the k-nearest neighbors algorithm and information theory to scale for real world applications. See our extensive paper describing these techniques on arXiv.
At the core of Howso is the concept of a Trainee, a collection of data elements that comprise knowledge. In traditional ML, this is typically referred to as a model, but a Trainee is original training data coupled with metadata, measured uncertainties and probabilities, details of feature attributes, with data lineage and provenance. Unlike traditional ML, Trainees are designed to be versatile so that after a single training instance (no re-training required!). They can:
- Perform classification on any target feature using any set of input features
- Perform regression on any target feature using any set of input features
- Perform online and reinforcement learning
- Perform anomaly detection based on any set of features
- Measure feature importance for predicting any target feature
- Identify counterfactuals
- Understand increases and decreases in accuracy for features and individual cases
- Forecast time series
- Synthesize data that maintains the same feature relationships of the original data while maintaining privacy
- And more!
Furthermore, Trainees are auditable, debuggable, and editable.
- Debuggable: Every prediction of a Trainee can be drilled down to investigate which cases from the training data were used to make the prediction.
- Auditable: Trainees manage metadata about themselves including: when data is trained, when training data is edited, when data is removed, etc.
- Editable: Specific cases of training data can be removed, edited, and emphasized (through case weighting) without the need to retrain.
Resources
General Overview
This repo provides the Python interface with Howso Engine that exposes the Howso Engine functionality. The Client objects directly interface with the engine API endpoints while the Trainee objects provides the python functionality for general users. Client functions may be called by the user but for most workflows the Trainee functionality is sufficient. Each Trainee represents an individual machine learning object with data set that can perform functions like training and predicting, while a client may manage the API interface for multiple Trainees.
Supported Platforms
Compatible with Python versions: 3.11, 3.12, 3.13, and 3.14.
Operating Systems
| OS | x86_64 | arm64 |
|---|---|---|
| Windows | Yes | No |
| Linux | Yes | Yes |
| MacOS | Yes | Yes |
Install
To install the current release:
pip install howso-engine
You can verify your installation is working by running the following command in your python environment terminal:
verify_howso_install
See the Howso Engine Install Guide for additional help and troubleshooting information.
Usage
The Howso Engine is designed to support users in the pursuit of many different machine learning tasks using Python.
Below is a very high-level set of steps recommended for using the Howso Engine:
- Define the feature attributes of the data (Feature types, bounds, etc.)
- Create a Trainee and set the feature attributes
- Train the Trainee with the data
- Call Analyze on the Trainee to characterize uncertainty and update it for inference
- Explore your data!
Once the Trainee has been given feature attributes, trained, and analyzed, then the Trainee is ready to be used for all supported machine learning tasks. At this point one could start making predictions on unseen data, investigate the most noisy features, find the most anomalous training cases, and much more.
Please see the User Guide for basic workflows as well as additional information about:
- Anomaly detection
- Classification
- Regression
- Time-series forecasting
- Feature importance analysis
- Reinforcement learning
- Data synthesis
- Prediction auditing
- Measuring model performance (global or conditional)
- Bias mitigation
- Trainee editing
- ID-based privacy
There is also a set of basic Jupyter notebooks to run that provides a complete set of examples of how to use Howso Engine.
License
Contributing
3 maintainers
Metadata
3 maintainers
Release files for howso-engine 68.2.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| howso_engine-68.2.1.tar.gz | 1.1 MB | Without provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| howso_engine-68.2.1-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 2.3 MB
Release files / howso_engine-68.2.1.tar.gz
| Download URL | howso_engine-68.2.1.tar.gz |
|---|---|
| Size | 1.1 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
68fcc9c3117ce09736a2ad9dd01d7072e7dd1f5a50b1bfa341bf9fef21e7bb01
|
|
BLAKE2b-256 checksum How to use checksums |
bfa5b9aa3a335551f101e772d2998ddcb39736ec2bf87d7a09e05395acd5cb95
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Release files / howso_engine-68.2.1-py3-none-any.whl
| Download URL | howso_engine-68.2.1-py3-none-any.whl |
|---|---|
| Size | 1.1 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
3c82dfc26ff090b1e106ddb155d391cda9ead512de549fa1acad00e7a4aa476f
|
|
BLAKE2b-256 checksum How to use checksums |
9d6f3ab1cdbfc735ef94c8dbf54a06948150d30d1f30926d0113e5e4991c88ab
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Security for howso-engine 68.2.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
