AbstrakDAMSMEL

Welcome to AbstrakDAMSMEL, and get ready for continuous optimisation.

Overview

AbstrakDAMSMEL (Abstrak-Directional Adaptive Metric Sampling Minimal Expected Loss), is a Python package for mathematical optimisation, specifically, continuous optimisations.

DAMSMEL (Directional Adaptive Metric Sampling Minimal Expected Loss) is an optimisation method for continuous optimisation, currently for unconstrained problems.

DAMSMEL uses adaptive metric sampling with exponentially decaying distance of adjacent points and by exploiting the minimal expected loss in the samples. In our paper (Purnawan and Adzkiya, 2024), we have shown DAMSMEL's convergence for convex optimisation landscapes. We have also performed empirical tests on DAMSMEL in several problems and have shown that DAMSMEL tends to provide better accuracies and reliability compared to gradient-based optimisations. We have also demonstrated DAMSMEL capacity as a machine learning model for medium scale regression problems in this notebook. All of our current tests on DAMSMEL can be observerd here.

Reference
Purnawan, R.; Adzkiya, D. (2024). Directional Adaptive Metric Sampling Minimal Expected Loss: A Continuous Optimisation Method. PREPRINT (Version 1). Research Square. DOI: 10.21203/rs.3.rs-5402563/v1

Importing AbstrakDAMSMEL

Follow the command below to import the modules into your notebook or local machine:

1. First, install the package using the following command:

!pip install git+https://github.com/rizalpurnawan23/AbstrakDAMSMEL.git

2. Then import damsmel module using the following command:

from damsmel import damsmel

There are two main Python classes in damsmel module, namely DAMSMELRegressor for machine learning regression implementations and DAMSMEL for general continuous optimisations, which can also be imported as follows:

from damsmel.damsmel import DAMSMELRegressor

from damsmel.damsmel import DAMSMEL

For ease of use, watch our tutorials on our Youtube channel.

Contributors

DAMSMEL is developed by Rizal Purnawan¹ and Dieky Adzkiya² as in independent research project.

While AbstrakDAMSMEL is an implementation of DAMSMEL into a Python package, developed by Rizal Purnawan¹ and Dieky Adzkiya².

¹ORCID: 0000-0001-8858-4036
²ORCID: 0000-0002-4718-2871
²Department of Mathematics, Institut Teknologi Sepuluh Nopember

License

AbstrakDAMSMEL is under MIT License. So, feel free to use it. We hope it could be helpful for everybody.

Get in Touch with Us