ICPOptimize
The Iterative Constrained Pathways Optimizer
ICP is a constrained linear model optimizer built with a focus on memory efficiency, flexibility, and solution interpretability.
Description
This repository contains implementations of the Iterative Constrained Pathways (ICP) optimization method, the ICP Rule Ensemble (ICPRE), linear classifier, regressor, and other methods. Currently, hinge, least-squares, and absolute-value loss modes are supported, with support for other loss functions planned. Coefficients can be constrained by sign or by arbitrary intervals. L1 & L2 norm constraints are also supported.
Further discussion about and motivation for the methods can be found on my blog:
nicholastsmith.wordpress.com/2021/05/18/the-iterative-constrained-pathways-optimizer/
Features
- Linear Classification using Hinge Loss
- Regression Support using L1 and L2 Penalties
- Arbitrary Interval Constraints
- L1 and L2 Coefficient Norm Constraints
- Useful Default Settings
- Support for DataFrames and Sparse Matrices
Installation
Install via PyPi:
pip install ICPOptimize
PyPi Project:
https://pypi.org/project/ICPOptimize/
Examples
Rule Ensemble Classifier
from ICP.Models import ICPRuleEnsemble
...
IRE = ICPRuleEnsemble().fit(A[trn], Y[trn])
YP = IRE.predict_proba(A)
Linear Model
from ICP.Models import ICPLinearRegressor
...
# Fit linear regressor with absolute loss and L_1 norm <= 10
ILR = ICPLinearRegressor(p='l1', L1=10.0).fit(A[trn], Y[trn])
YP = ILR.predict(A)
Further examples are available on the ICPExamples GitHub page:
1 maintainer
Release files for ICPOptimize 2.4
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| ICPOptimize-2.4.tar.gz | 241.2 kB | Details |
Release files / ICPOptimize-2.4.tar.gz
| Download URL | ICPOptimize-2.4.tar.gz |
|---|---|
| Size | 241.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
464d1fdb8094a45ea27b0c2882617991a7f8806af62a4a046daa61682d505636
|
|
BLAKE2b-256 checksum How to use checksums |
b2cf8a798b4697ce12f5f8d61772c5000119edd9c2f4f00034a03e9d3b98a45a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/3.7.1 importlib_metadata/4.11.3 pkginfo/1.8.2 requests/2.27.1 requests-toolbelt/0.9.1 tqdm/4.64.0 CPython/3.9.12
|
