SAP digs deeper roots for machine learning with 'foundation' expansion

German enterprise software company SAP has used its TechEd Europe 2017 conference and exhibition to detail news of its SAP Leonardo Machine Learning Foundation (which is based on the SAP Cloud Platform) and the foundation’s current expansion.

The firm uses the term ‘foundation’ in this case to describe a technology, a software system and a team or technologists putting forward a platform — so this is not a foundation as in a charitable foundation for donkeys, for example.

In terms of working with machine learning in general, SAP earlier this year updated its Hybris Marketing Cloud with machine learning technologies for facial recognition and other Internet of Things (IoT) augmentations.

Current expansions within the SAP Leonardo Machine Learning Foundation see three new three major capabilities added:

  • ready-to-use services
  • retrainable services
  • a bring-your-own model

Customers, partners and developers can now rely on additional ready-to-use services such as optical character recognition (OCR) and multidimensional time series forecasting to predict the optimal price for a product, for example.

Additionally, SAP Leonardo Machine Learning Foundation now allows customers and partners to tailor generic models with company-specific data. The new image classification service can be retrained to recognise customer-specific objects such as products, components or spare parts.

Predefined, or retrained

Using predefined services or retrained models requires only minimal machine learning experience.

“SAP Leonardo Machine Learning Foundation puts digital intelligence into the hands of all customers, partners and developers – whether they are just getting started or already have dedicated data science teams. We aim to deliver offerings that are the easiest to use and have the most impact on companies,” said Markus Noga, head of machine learning, SAP.

Customers and partners with their own data science teams can now deploy and run their custom models based on TensorFlow models from Google LLC on SAP Leonardo Machine Learning Foundation. These advanced users can benefit from interoperability with SAP Cloud Platform.

As SAP has explained before no, the intention with SAP Leonardo Machine Learning Foundation is that application developers with no AI expertise have been able to use ready-to-use services to jumpstart the development of what has been called more intelligent applications.

These services are pre-trained to work on real data, readily available on SAP Cloud Platform and accessible via standard (RESTfull) APIs… and this makes them fit (so says SAP) for easier consumption.

A working example

SAP pre-trained model for image recognition can classify a variety of different objects. It can tell apart anything from cars to people, to trees. However, if a car manufacturer wants to facilitate visual shopping by accurately matching a given picture to one of its products, it needs to teach the foundation’s Image Recognition service to recognise different car models. In the same way, the wallpaper retailer mentioned before had to teach the image recognition model how to recognize different kinds of surface textures.

Image credit SAP