Mathematics Developer in Edmonton, AB, Canada

SKSurrogate is a suite of tools which implements surrogate optimization for expensive functions based on scikit-learn. The main purpose of SKSurrogate is to facilitate hyperparameter optimization for machine learning models and optimized pipeline design (AutoML).

The version of the surrogate optimization implemented here heavily relies on regressors. A custom regressor based on Hilbert Space techniques is implemented, but all scikit-learn regressors are accepted for optimization.

Finding an optimized pipeline—based on a given list of transformers and estimators—is a time-consuming task. A version of evolutionary optimization has been implemented to reduce its time in lieu of global optimality.

Irene is a Python package that aims to be a toolkit for global optimization problems that can be realized algebraically. It generalizes Lasserre's Relaxation method to handle theoretically any optimization problem with a bounded feasibility set. The method is based on solutions of generalized truncated moment problem over commutative real algebras.

Given inventory data of multiple (interacting) commodities from stock with limited but variable capacity, provide insight on:
1. Estimating future required capacity for each item based on a certain terminal segment of data,
2. Future cost estimation for each item,
3. How the trends of individual items would change, assuming a trend change at given times (in future) for some items?
4. Given a budget limit, how should the trends change to make sure a non-negative residual?

This package was originally written to solve systems of integro-differential equations via collocation method in an arbitrary number of variables. The current implementation of the method is based on a finite-dimensional orthogonal system of functions. Therefore additional modules were required to achieve this goal.