The Alchemite™ 2023 Autumn Release is a milestone release of the advanced machine learning software from Intellegens, bringing together the many new features delivered over the past six months. New features fall under three main themes.
Optimised usability for larger, denser datasets – a range of user interface and algorithmic improvements make it easier to work with the large datasets often encountered in chemical, materials, and life science R&D.
More flexible and powerful exploration of design space – specify complex constraints, set up new experimental programmes, and get the guidance that you need to focus-in on the best formulations, materials, and chemicals.
Applied machine learning for real R&D teams – onboard new users faster and benefit from new capabilities to support collaboration and communication of results.
What’s new?
For larger, denser datasets
Tuning the algorithm
The new complete columns feature auto-identifies datasets where descriptor data is complete for every column, enabling the use of a faster algorithm for this common experimental scenario.
A slick user experience
Speedups in the data explorer and analytics make interacting with and analysing larger datasets a quicker, easier experience.
Revved-up analytics
The dimensionality reduction plot can now be used for datasets up to 10,000 rows, aiding understanding of large, multi-dimensional datasets. More options to sort and filter importance plots also eases analysis of larger datasets.
For exploring design space
Results you can count on
Bespoke property hyperparameters, an algorithmic optimisation, delivers greater accuracy in predictions. The estimates of uncertainty when extrapolating beyond a dataset or exploring gaps in the data have also been improved, providing better guidance for decision-making in design of experiments.
DOE made easy
Calculated columns is a new feature enabling the user to define relationships between columns in a dataset, making it easier to include domain knowledge when setting constraints or analysing results. A new user interface for specifying the design space when setting up an adaptive design of experiments project aids design of experiments (DOE) when you have to consider tens or hundreds of factors.
If you have little (or no) data…
Did you know that, even if you have no data, Alchemite™ can get you started by recommending which experiments to do first? The ‘initial experiments’ tool is more useful than ever, thanks to usability improvements and the inclusion of categorical data types and constraints.
For real R&D teams
Supporting communication and collaboration
Automated generation of PDF summary reports makes it simple to share project results with colleagues, partners, and managers. And shareable permalinks mean you can even enable collaborators to directly view analytics results and experimental suggestions within the Alchemite™ platform.
Faster onboarding
A host of small user interface and workflow improvements, made based on direct feedback from users, not only speed things up for these existing users, but reduce the barriers for onboarding new users – share the benefits of Alchemite™ across your team!
