Data

Tools

Indexes

Applications

Experiments

Open Source Science

ccl_tool

Chance Constraint Learning Framework

https://github.com/antonioalcantaramata/ccl_tool

Science Score: 44.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ○
    Academic publication links
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (10.4%) to scientific vocabulary
Last synced: 9 months ago · JSON representation ·

Repository

Chance Constraint Learning Framework

Basic Info
  • Host: GitHub
  • Owner: antonioalcantaramata
  • Language: Python
  • Default Branch: main
  • Size: 349 KB
Statistics
  • Stars: 6
  • Watchers: 1
  • Forks: 2
  • Open Issues: 0
  • Releases: 0
Created almost 4 years ago · Last pushed almost 3 years ago
Metadata Files
Readme Citation

README.md

ccl_tool

Chance Constraint Learning (CCL): a complete framework for chance constraint learning is developed to allow practitioners to solve chance constraints and also adding probabilistic guarantees over learned constraints.

Note: Most of the code for point estimation is derived from the work in https://github.com/donato-maragno/OptiCL

Usage

Initialization

After cloning or downloading all the .py files in your working environment, you can initialize the ccl_tool.

from CCL import CCL ccl_tool = CCL(X, y, methodology, p_model, q=None, M_super=None, side=None)

  • X and y represents the independent and dependent variable in our dataset, respectively.
  • methodology is one of ['point', 'quantile', 'superquantile']
  • p_model is one of ['lr', 'svm', 'tree', 'rf', 'gbm', 'nn']
  • q is used to fit the quantile when using 'quantile' or 'superquantile' as methodology, and ranges in (0,1)
  • Msuper is the number of quantiles used to approximate the superquantile (only when pmodel is not a tree-based method)
  • side in ['left', 'right'] indicates the side of the quantile to use when computing the superquantile

Training

model = ccl_tool.train()

We save the trained p_model. In most cases, a hyper-parameter selection is applied.

Constraint learning

cons = ccl_tool.constraint_build(model)

Linear (sometimes with binary variables) constraints are generated to represent the p_model and be embedded within our MIO problem.

Constraint embedding

conceptual_model= init_conceptual_model(cost_p) ccl_tool.const_embed(conceptual_model, cons, 'y_name', lb, ub)

conceptualmodel represents our initialized MIO problem in Pyomo. Generated constraints (cons) are embedded in the model (conceptualmodel). A name for the learned variables is needed ('y_name'), and an additional lower and/or upper bound for the chance constraint (lb, ub).

Citation

If you use this code, you can cite it in you work as:

@article{alcantara2023data, title={On data-driven chance constraint learning for mixed-integer optimization problems}, author={Alc{\'a}ntara, Antonio and Ruiz, Carlos}, journal={Applied Mathematical Modelling}, volume={121}, pages={445--462}, year={2023}, publisher={Elsevier} }

Owner

  • Name: Antonio Alcántara
  • Login: antonioalcantaramata
  • Kind: user
  • Location: Madrid
  • Company: Universidad Carlos III de Madrid

Citation (citation.cff) 

cff-version: 1.2.0
message: "If you use this software, please cite it as below."
authors:
- family-names: "Alcántara"
  given-names: "Antonio"
title: "ccl_tool"
version: 1.0.0
date-released: 2022-07-05
url: "https://github.com/antonioalcantaramata/ccl_tool"

GitHub Events

Total
  • Watch event: 2
  • Fork event: 1
Last Year
  • Watch event: 2
  • Fork event: 1