https://www.meetup.com/Technologietrends-und-Innovation-fur-die-Praxis/ Introducing AutoAI with IBM Studio

Benedikt Klotz [email protected] linkedin.com/in/benediktklotz

IT Architect, IBM Hybrid Cloud

October 3rd 2019 AutoAI definitions from various sources

Automated (AutoML) is the process of automating the end-to-end process of applying 1 machine learning to real-world problems.

AutoML is the capability to automatically ingest, clean, transform, and model with hyperparameter 2 optimization

3 Auto ML services provide machine learning at the click of a button, or, at the very least, promise to keep algorithm implementation, data pipelines, and code, in general, hidden from view.

Quite simply, it is the means by which your business can optimize resources, encourage collaboration 4 and rapidly and dependably distribute data across the enterprise and use that data to predict, plan and achieve revenue goals. AutoAI definitions from various sources

Automated machine learning (AutoML) is the process of automating the end-to-end process of applying 1 machine learning to real-world problems. Wikipedia

AutoML is the capability to automatically ingest, clean, transform, and model with hyperparameter 2 optimization IBM

3 Auto ML services provide machine learning at the click of a button, or, at the very least, promise to keep algorithm implementation, data pipelines, and code, in general, hidden from view. Blogger

Quite simply, it is the means by which your business can optimize resources, encourage collaboration 4 and rapidly and dependably distribute data across the enterprise and use that data to predict, plan and achieve revenue goals. Blogger IBM Watson operationalizes AI across the enterprise

Organize and Build Deploy and run Operate trusted AI Govern data

IBM TOOLS IBM AI RUNTIME Manage AI at scale

Watson Knowledge Catalog Watson Studio Watson Machine Learning Watson OpenScale

3RD PARTY IDE & 3RD PARTY RUNTIMES Business KPIs Data Profiling FRAMEWORKS Amazon Web Services Validation and Feedback Spark ML Microsoft Azure ML Accuracy Quality and Lineage Scikit-learn SPSS Modeler Fairness and Explainability Keras Data Governance Custom (Kubernetes etc.) Inputs for Continuous Evolution Pytorch Automated Anomaly and Drift Caffe2 … detection

Consume AI Build AI Run AI Organize Data for AI Business user Data Scientist App Developer Data Engineer 5 Case for AI Automation: AI Workflow’s Bigger & More Complex

Ground Truth Gathering Model Data Cleansing Improvement

Runtime Feature Monitoring Engineering AI Lifecycle Management

Model Model Selection Deployment

Majority spent on Model Parameter Validation Optimization data wrangling! Ensemble

Source: https://www.kaggle.com/paultimothymooney/2018-kaggle-machine-learning-data-science7 -survey Why care about Automation?

Coming up with features is difficult, time-consuming, requires expert knowledge. "Applied machine learning" is basically . -Andrew Ng

...some machine learning projects succeed and some fail. What makes the difference? Easily the most important factor is the features used. - Pedro Domingos

The algorithms we used are very standard for Kagglers. […] We spent most of our efforts in feature engineering. [...] -Xavier Conort

Feature Engineering is essential, difficult and costly. The Gap: Data scientists acknowledge lack of expertise

Data Scientist Jobs Explosion • ~50% of Data Scientist respondents on Kaggle said they had less than 2 years of experience on ML methods. Same for coding experience

• Significant percentage uptick in AI roles being adopted in vertical industries (LinkedIn 2018 jobs report) (again more 2016 2017 2018 domain/application knowledge, less ML and CS)

How long have you been How long have you used Machine Learning writing code methods

https://www.kaggle.com/paultimothymooney/2018-kaggle-machine-learning-data-science-survey https://economicgraph.linkedin.com/research/linkedin-2018-emerging-jobs-report https://economicgraph.linkedin.com/research/LinkedIns-2017-US-Emerging-Jobs-Report AI for AI: Livecycle of Automation of AI Development and Operation

AI Designing AI AI Optimizing AI AI Governing AI

Lifecycle management Neural network AI pipeline optimization Monitoring AI outcome with architecture and search Decision optimization trust and explainability IBM’s Portfolio for Automation of AI Development

May 2019

Transfer Learning AutoAI Experiments & Neural Network Search Pipeline optimization • Transfer knowledge learning in • Just bring data and one system to • Auto clean data, engineer automatically generate a apply to a different domain features, and complete HPO custom deep neural network to find the optimal end to end through searching the best • Featured in Watson Visual pipeline architectures for the input data Recognition or NLP Services, available via Watson Studio • AutoAI as a feature of • NeuNetS as a feature of Watson Watson Studio Studio, available in Open Beta What does AutoAI with IBM Watson Studio do?

• Integrated with Watson Studio and Watson Machine learning

• Automatically ingest, clean, transform, and model with hyperparameter optimization

• Training feedback visualizations provide real-time results to see model performance

• One-click deployment to Watson Machine Learning

© 2018 IBM Corporation How does it work?

AutoAI Pipeline Optimization

Model Feature Raw Labeled Prep HPO HPO Ensembling Data Set selection Engineering

Finds best Finds top-K HPO on Finds best data HPO on Computes preprocessing estimators selected transformation estimator ensemble imputation / estimator sequence after Feature predictions encoding and Engineering based on scaling strategies pipeline predictions

© 2018 IBM Corporation How does it work?

Watson Studio

AutoAI Experiment

Import/Add Raw Model Feature Prep HPO HPO Ensembling Project Data Labeled selection Engineering Data Set

Train Model saved Infrastructure to WML

Deploy

REST prediction App hosted API model Artifact Repository

Watson Machine Learning

User Benefits of using AutoAI

Build models faster Automate data preparation Find signal from noise and model development Auto-feature engineering makes it easy to extract more predictive power from your data Jump the skills gap No coding? No problem – get Rank and explore models started with a couple clicks Quickly compare candidate pipelines to find the best model for the job

Discover more use Ready, set, deploy cases Pipelines generated with AutoAI can Supercharge collaboration with be deployed to REST APIs with one AI everywhere to disrupt and click transform Get Started Today

AutoAI:

• Accelerate • Automate • Optimize

Experience AutoAI: https://www.ibm.com/demos/collection/IBM-Watson-Studio-AutoAI/

AI Designing AI NeuNetS: A Breakthrough in AI Designing AI

In a few hours Overnight / day light weight Weeks to a month

23 | © 2018 - IBM Corporation Copyright | IBM Confidential Design Neural Networks is challenging

• Highly skilled researchers/data-scientists are needed to hand-craft custom neural networks • Hand-crafting complex networks is time-consuming, error prone, and does not scale with time and resources • Neural networks continue to grow in size and complexity

Think 2019 / DOC ID / Month XX, 2019 / © 2019 IBM 24 Corporation