End-to-End Implementation with SAP's Data Science and Machine Learning Platform Webcast Recap

[TammyPowlas]

Source: SAP

This was a SAP User Group webcast from last month. The product was announced at SAPPHIRENOW earlier this month.

Abstract: (source: SAP User group)

In this session we will present a preview of SAP’s Data Science and Machine Learning Platform. We will showcase how to use this new platform to implement a Machine Learning project end-to-end.

The webcast will moreover cover machine learning scenarios which address standard business problems. A machine learning scenario consists of pipelines for training, validation, and the final application. A pipeline for training can entail several machine learning tasks, and each task can use assets like a data set or a neural network.

Source: SAP

Legal Disclaimer applies



Source: SAP

SAP is developing technologies to develop machine learning services and data science platform where you can build and maintain, and manage your own machine learning scenarios

Approach, architecture, use cases

Focus of this session is SAP Data Intelligence



Source: SAP

Machine learning was released 2 years ago

Services in areas of image, speech and text

Can train TensorFlow models externally

Based on SAP Cloud Platform



Source: SAP

Various image services at the moment – classification, face detection

Own application and customers

Product ID, know product ID, use image service to take a picture of product and system can tell you what it is



Source: SAP

Speech area – open, partnering with Google

Partnering with Google, using Google Services



Source: SAP

Text feature extraction services, language services



Source: SAP

Use cases for machine learning



Source: SAP

Platform to manage machine learning (ML) use cases and scenarios end to end and scale them

What do you need to do a ML scenario

First, identify business need

Big topic is data prep, manage data, streams lined up, annotate data, prepare data to train ML model and algorithm

Need to have services to create ML models and train with data you have prepared.

Access algorithms easily from platform

Challenge is today – customers may have good ML scenarios, data scientists, but when it comes to deployment of ML in production, may have some challenges, not specific technique - look at TensorFlow

Retrain models, versioning and auditable for external requirements

SAP is "not the new AI company"

Providing end to end platform to manage ML scenarios and get them into production

Looking at TensorFlow, R, state of art libraries

Unify ML and Data Science platform

One end to end platform



Source: SAP

Combining various solutions in SAP into one offering



Source: SAP

One data science platform, ML IDE

Ops team, data scientist, make use of one product

Data scientist, developer, can make use of one product, one IDE



Source: SAP

Need IT ops for ML scenarios

Data scientist solve ML problem



Source: SAP

Advantage of platform is it is scalable, manage, deploy, ML scenarios

Can manage scenarios in parallel

Want to automate training, maintenance, retirement of models

Platform is open to language and framework

Bring your own ML models

Deploy in AWS/Google data centers

End to end management of scenarios

Embed in SAP applications

SAPPHIRENOW – SAP will launch new SAP Data Intelligence platform, which is the combination of various products mentioned

Links:

Recording

PDF​

What do you think?