The Optimization Firm Introduces ALAMO: Machine Learning Tool for Building Interpretable Models from Data

The Optimization Firm this summer officially launched ALAMO, a machine learning tool that makes it easy to build accurate and simple models that require minimal data and can impose physics and chemistry constraints wherever possible. ALAMO offers a spreadsheet-like GUI, allowing any user to build their models with just a few clicks. Coming from an undisputed leader in mathematical programming and analytics, ALAMO offers several input/output formats, graph options, and the ability to systematically investigate the effect of a vast number of basis functions, and works with familiar simulators, including Python, Excel, and MATLAB. ALAMO leverages BARON, The Optimization Firm’s powerful global optimization framework, and the cutting-edge technology behind ALAMO’s modeling process, including constrained regression, adaptive sampling, and low discrepancy initialization.

Solving large-scale problems quickly and efficiently has always been core to The Optimization Firm’s mission of developing and implementing novel algorithms. ALAMO makes it possible for everyone, independent of skill level, to build meaningful models.

Recommended AI News: Pega Launches Pega Process Extender for Salesforce Lightning on Salesforce App Exchange and Pega Marketplace

“The world is awash in data, and it’s our mission to support anyone building models from data,” said Dr. Nick Sahinidis, co-founder and CEO of The Optimization Firm. “We are excited to launch ALAMO as a simple, spreadsheet-like tool that unifies the sampling, predicting, and optimization process so users can easily build accurate, interpretable models.”

Scientists and engineers routinely build models from data using machine learning techniques, such as linear regression, lasso, and neural networks. However, the downside is that some methods make oversimplifying assumptions about the models, while others achieve accuracy but require vast amounts of input data and return models that are highly complicated and difficult to interpret.

Recommended AI News: Tyson & Blake makes Follow-on Investment in StrongRoom AI “SRAI”

Some researchers think machine learning could offer a better way to build models from data. At Carnegie Mellon University in 2012, Dr. Nick Sahinidis described a method that can achieve the accuracy of neural networks using the smallest possible number of measurements while applying physical constraints. He, along with his then-Ph.D. student, Dr. Alison Cozad, called it Automated Learning of Algebraic Models (ALAMO).

Today, ALAMO is distributed by The Optimization Firm and has been used by over 500 engineers and data scientists in industries including automation, chemicals, computer software, pharmaceutics, and manufacturing.

“With ALAMO, we are meeting the demands of data-driven decision-making,” said Dr. Yi Zhang, Optimization Research Scientist at The Optimization Firm. “Providing interpretability and accuracy, ALAMO is a necessary tool for people and companies trying to solve highly nonlinear regression and classification problems in complex environments and systems.”

Recommended AI News: Datanomix Announces Automated Downtime Insights

[To share your insights with us, please write to sghosh@martechseries.com]