In a previous blog, I covered the reasons why SAP is bringing the Intelligent Enterprise vision to the Market, in order to help companies tackle the digitalization challenge.
The purpose of this new blog is to try and explain the basic concepts of Machine Learning, and to give an overview of how it is being leveraged in SAP’s product portfolio.
One key aspect of the Intelligent Enterprise is infusing Machine Learning algorithms into our Enterprise Software business processes. Machine Learning algorithms are nothing new in themselves, but they’ve been taking a huge importance lately because we’re at a time when the technology has become powerful enough to leverage them in an efficient and economical way, and the huge quantity of data generated since the dawn of the digital economy is now available to train them to produce quality outcomes.
What is Machine Learning?
The goal of a Machine Learning algorithm is to determine a mathematical model, that we will call the function f(), in y = f(x) where x represents any real-world observation and y represents some evaluation, recognition, or result caused by those x.
There are three major Machine Learning algorithm types: Supervised Learning, Semi-Supervised Learning, and Unsupervised Learning.
To add another layer of complexity, the function f() we’re intending to define is typically influenced by very many variables, with possible dependencies between one another: y = f(x1, x2, x3, … , xn).
The Machine Learning algorithm builds a mathematical model f() from the training data, with the ultimate goal of making statistical predictions or decisions without having been explicitly programmed to do so.
“Training” a model then means determining “correct” values (y) for all (xn) from examples from the training data set. For example, in supervised learning, a machine learning algorithm builds a model by examining the many labeled data (xn) and attempting to find a model that minimizes the loss. The loss is the number indicating how bad the model's prediction is on a single example (a bad (y)). If the model's prediction was perfect, the loss would be zero; otherwise, the loss is higher. The goal of training a model is to find a set of (xn) that have low loss, then inferring a correct (y).
Please find in the Figure below the mechanism for training a machine learning model:
This is why the biggest the training data set is, and the best quality the data has, the more accurately we can converge to the function f(), and the most accurate the predictions will be.
Hence for machine learning to be successful, we need to prepare huge data sets of high-quality data.
Machine Learning at SAP
SAP’s vision for machine learning encompasses:
1- Intelligent business processes, embedded into the ERP S/4HANA and the LOB SaaS tools (SuccessFactors, Ariba, C/4HANA, Concur, etc.);
2- The intelligent Digital Platform, which is powering chatbots, virtual assistants, robotic process automation and provides a vast library of Machine Learning algorithms developed by SAP and by our partners, such as Google’s TensorFlow.
3- Finally, SAP HANA, as a foundation, provides all the data governance engines necessary to prepare, govern, ensure the quality, and treat in real time the data necessary to train its own predictive analysis engines, and/or other machine learning algorithms available from the SAP Cloud Platform.
To conclude this overview, I will be introducing SAP’s key offerings around Machine Learning.
SAP Leonardo Machine Learning Foundation:
SAP Leonardo Machine Learning Foundation is an SAP Cloud Platform-based platform for machine learning enabling simple consumption and providing a tight integration with the SAP backend. Its four main capabilities are:
1- Easy to consume ML content targeted at non-ML experts allowing to deploy and run your own machine learning models or tune existing models with your own data.
2- Image processing services to enable automatic pattern detection
3- Text processing service that analyzes natural language content stored in documents, Web sites, or e-mails, and reveals its meaning.
4- Speech processing service that converts speech to text and synthesizes speech to power digital assistants or voice-controlled apps.
SAP Intelligent Robotic Process Automation:
SAP Intelligent Process Automation (IPA), also based on SAP Cloud Platform, combines Robotic Process Automation and Machine Learning into one integrated automation offering:
▪ ML to “think” and handle unstructured processes/data and to improve pure rule-based decision engines.
▪ RPA to “act”, esp. execute business processes across systems and let key users create their own bots.
The purpose of IPA is to accelerate the digital transformation of Business Processes by automatically replicating tedious actions that present no added value. It is about Business Process automation across applications and systems.
SAP Conversational AI
SAP Conversational AI is an SAP Cloud Platform-based platform to develop, deploy, & monitor Conversational AI applications. It comes with powerful natural language processing (NLP) technology, so you can build bots that truly understand people – quickly, and easily.
The bot is perfectly fluent in English, French, Spanish and German, but also offer standard functionalities in 15 other languages.
Conclusion
The utilization of Machine Learning to improve and automate business processes is only limited by the power of our imaginations. This is ultimately what Intelligent Enterprise is all about. Beyond the actual technology, the goal of all these tools is to help you and create value for your business.
The purpose of this new blog is to try and explain the basic concepts of Machine Learning, and to give an overview of how it is being leveraged in SAP’s product portfolio.
One key aspect of the Intelligent Enterprise is infusing Machine Learning algorithms into our Enterprise Software business processes. Machine Learning algorithms are nothing new in themselves, but they’ve been taking a huge importance lately because we’re at a time when the technology has become powerful enough to leverage them in an efficient and economical way, and the huge quantity of data generated since the dawn of the digital economy is now available to train them to produce quality outcomes.
What is Machine Learning?
The goal of a Machine Learning algorithm is to determine a mathematical model, that we will call the function f(), in y = f(x) where x represents any real-world observation and y represents some evaluation, recognition, or result caused by those x.
There are three major Machine Learning algorithm types: Supervised Learning, Semi-Supervised Learning, and Unsupervised Learning.
- Supervised Learning works with labeled data, meaning finding out the equation y = f(x) to predict/infer some specific result/outcome (y) based on real-world observations (x)
by analyzing training data. - Semi-supervised learning works with some labeled data and a lot of unlabeled data, meaning finding out the equation y = f(x) even if the result values (y) are only known only for a limited set of observations (x), e.g. in a sampling scenario.
- Unsupervised learning works with only (x) input data, meaning finding out hidden structures or relationships between the observations (x). The variable (y) being automatically generated from this analysis.
To add another layer of complexity, the function f() we’re intending to define is typically influenced by very many variables, with possible dependencies between one another: y = f(x1, x2, x3, … , xn).
The Machine Learning algorithm builds a mathematical model f() from the training data, with the ultimate goal of making statistical predictions or decisions without having been explicitly programmed to do so.
“Training” a model then means determining “correct” values (y) for all (xn) from examples from the training data set. For example, in supervised learning, a machine learning algorithm builds a model by examining the many labeled data (xn) and attempting to find a model that minimizes the loss. The loss is the number indicating how bad the model's prediction is on a single example (a bad (y)). If the model's prediction was perfect, the loss would be zero; otherwise, the loss is higher. The goal of training a model is to find a set of (xn) that have low loss, then inferring a correct (y).
Please find in the Figure below the mechanism for training a machine learning model:
This is why the biggest the training data set is, and the best quality the data has, the more accurately we can converge to the function f(), and the most accurate the predictions will be.
Hence for machine learning to be successful, we need to prepare huge data sets of high-quality data.
Machine Learning at SAP
SAP’s vision for machine learning encompasses:
1- Intelligent business processes, embedded into the ERP S/4HANA and the LOB SaaS tools (SuccessFactors, Ariba, C/4HANA, Concur, etc.);
2- The intelligent Digital Platform, which is powering chatbots, virtual assistants, robotic process automation and provides a vast library of Machine Learning algorithms developed by SAP and by our partners, such as Google’s TensorFlow.
3- Finally, SAP HANA, as a foundation, provides all the data governance engines necessary to prepare, govern, ensure the quality, and treat in real time the data necessary to train its own predictive analysis engines, and/or other machine learning algorithms available from the SAP Cloud Platform.
To conclude this overview, I will be introducing SAP’s key offerings around Machine Learning.
SAP Leonardo Machine Learning Foundation:
SAP Leonardo Machine Learning Foundation is an SAP Cloud Platform-based platform for machine learning enabling simple consumption and providing a tight integration with the SAP backend. Its four main capabilities are:
1- Easy to consume ML content targeted at non-ML experts allowing to deploy and run your own machine learning models or tune existing models with your own data.
2- Image processing services to enable automatic pattern detection
3- Text processing service that analyzes natural language content stored in documents, Web sites, or e-mails, and reveals its meaning.
4- Speech processing service that converts speech to text and synthesizes speech to power digital assistants or voice-controlled apps.
SAP Intelligent Robotic Process Automation:
SAP Intelligent Process Automation (IPA), also based on SAP Cloud Platform, combines Robotic Process Automation and Machine Learning into one integrated automation offering:
▪ ML to “think” and handle unstructured processes/data and to improve pure rule-based decision engines.
▪ RPA to “act”, esp. execute business processes across systems and let key users create their own bots.
The purpose of IPA is to accelerate the digital transformation of Business Processes by automatically replicating tedious actions that present no added value. It is about Business Process automation across applications and systems.
SAP Conversational AI
SAP Conversational AI is an SAP Cloud Platform-based platform to develop, deploy, & monitor Conversational AI applications. It comes with powerful natural language processing (NLP) technology, so you can build bots that truly understand people – quickly, and easily.
The bot is perfectly fluent in English, French, Spanish and German, but also offer standard functionalities in 15 other languages.
Conclusion
The utilization of Machine Learning to improve and automate business processes is only limited by the power of our imaginations. This is ultimately what Intelligent Enterprise is all about. Beyond the actual technology, the goal of all these tools is to help you and create value for your business.
- SAP Managed Tags:
5 Comments
