An emerging trend in AI is the availability of technologies in which automation is used to select a best-fit model, perform feature engineering and improve model performance via hyperparameter optimization. This automation will provide rapid-prototyping of models and allow the Data Scientist to focus their efforts on applying domain knowledge to fine-tune models. This course will take the learner through the creation of an end-to-end automated pipeline built by Watson Studio’s AutoAI experiment tool, explaining the underlying technology at work as developed by IBM Research. The focus will be on working with an auto-generated Python notebook. Learners will be provided with test data sets for two use cases.
Content
Building a rapid prototype of Watson Studio AI - Describe the benefits of AutoAI for rapid prototyping - Identify implementations of AutoAI - Become familiar with the Watson Studio platform - Build rapid prototypes using Watson Studio AutoAI - Generate a Python notebook of the prototype with one click Automated Data Preparation and Model Selection - Evaluate the data preprocessing steps for the use cases - Refine data preprocessing using the AutoAI-generated Python notebook - Examine the model selection outcome for use cases - Refine the Python notebook to make changes to the selected model Automated Feature Engineering and Hyperparameter Optimization - Explain how the Cognito algorithm can save time by automating feature engineering - Evaluate the automated feature engineering performance for the use cases - Describe several strategies for HPO in order of increasing soistication - Observe how changes to the model hyperparameters in the Python notebook affect the prototype's performance Evaluation and Deployment of AutoAI-generated Solutions - Evaluate the prototype for further development or deployment based on calculated performance metrics - Deploy the prototype using Watson Machine Learning