Democratization of Machine Learning Models
Altair EmployeeReusing Machine Learning models for automation usually faces difficulties when integrating diverse ML dependencies, to a large extent due to the lack of a common framework. In most cases, the ML models are developed for specific use cases. Others such as colleagues may also be interested in the models for the very same use cases or the author themselves may want to use them in different applications. Hence, to mitigate duplicate efforts, ML models can be distributed via some smart packaging methods. This blog will illustrate the capabilities of Altair HyperStudy in packaging ML models into portable containers which can be easily shared and can be integrated into various applications
Machine learning models are gaining more and more traction in almost every field imaginable and what is becoming more prominent as well as their accuracy is their portability. It may be valuable to make a model accessible to others and deployable in platforms other than its original one in where it is created.
There are numerous ways of making ML models portable and probably one of the most common examples is pickling models in python to be later deployed elsewhere. Once a model is saved, it can be shared with whoever wants to take advantage of it whether for simple trade-off studies or automation projects or co-simulation purposes. Altair HyperStudy provides three options for packaging ML models:
- HyperStudy Pyfit, a proprietary format (*. pyfit) by HyperStudy. The pyfit files can be used as an evaluation engine in Altair products like Activate and Compose.
- Excel Spreadsheet, ML models are embedded in a spreadsheet which can easily be shared with others for quick trade-off studies.
- Functional Mock-up Unit, HyperStudy supports Functional Mock-up Interface and can save models in (*.fmu) format. This format brings more freedom as it is also supported by variety of both Altair and third-party platforms.
