Overview
The course on "Automated Machine Learning" addresses the challenge of designing well-performing Machine Learning (ML) pipelines, including their hyperparameters, architectures of deep Neural Networks and pre-processing. Future ML developers will learn how to use and design automated approaches for determining such ML pipelines efficiently. The course is designed either to be taken as a MOOC or can be offered by universities in a Blended Learning format with face-to-face and online phases.
Which topics will be covered?
-
In Hyperparameter Optimization, the hyperparameter settings of a given Machine Learning algorithm are optimized to achieve great performance on a given dataset.
-
In Neural Architecture Search, the architecture of a Neural Network is tuned for its predictive performance (or in addition inference time or model size) on a given dataset.
-
As AutoML optimizers, approaches such as Bayesian optimization, evolutionary algorithms, multi-fidelity optimization and gradient-based optimization are discussed.
-
Via Dynamic & Meta-Learning, useful meta strategies for speeding up the learning itself or AutoML are learned across datasets.
What will I achieve?
By the end of the course, you‘ll be able to…
-
identify possible design decisions and procedures in the application of ML.
-
evaluate the design decisions made.
-
implement efficient optimizers for AutoML problems, such as hyperparameter optimization and neural architecture search.
-
increase the efficiency of AutoML via a multitude of different approaches.
Which prerequisites do I need to fulfill?
-
Basics in Machine Learning (ML) and Deep Learning (DL)
-
First experiences in the application of ML & DL
-
Python or R as programming language
-
Recommended but optional: Basics of Reinforcement Learning