Engineering Applications of Machine Learning - Statistical Solutions by Creascience

Artificial Intelligence (AI) has become a prominent sale argument no matter whether you want to buy a simple toaster or an advanced assembly line robot. The purpose of this seminar is to demystify the jargon used and to provide attendees with the tools needed to evaluate the value of AI-based solutions in the context of engineering applications.

For this purpose, the focus is put on Machine learning (ML), which is the main building block of AI solutions. From its principles, paradigms and algorithms that are dramatically changing data analysis methods to its real-life implementations, ML is dissected to allow participants to get ideas on how they can implement by themselves ML solutions, whether it is to solve complex research problems or to implement efficient QC procedures. Furthermore, the workshop aims at offering a series of key questions to ask vendors of AI-based solutions to determine whether their product suits your needs.

Course Outline

The recommended duration for this course is 1 online session(s).

This 2.5h workshop will go through the following topics:

Introduction: AI is everywhere. But what is the link with ML?

Limitations of classical data analysis tools
- Linearity assumption and its weak alternatives
- Dealing with large datasets
- Predictive ability of models

ML new paradigms and alternative solutions
- A brief historical background - the rise of computing power
- Model specifications - from tree-based algorithms to deep learning
- Selecting the best predictive model: crossvalidation

Intermission: the shift in vocabulary between data analysis and ML: demystifying the jargon

The limitations of machine learning
- The need for appropriate data: bias vs. representativity
- Data preparation: data engineering should never be overlooked
- The shrinkage paradox and its potential consequences

ML and AI in practice - Putting the pieces together
- Preparing data
- Selecting the most appropriate model/algorithm
- Fitting the model(s)
- Quantifying the models accuracy
- Making the final selection and fine-tuning
- Deploying the model

Case-studies in engineering
- A few examples using historical data
- Implementations for QC purposes
- Using third-party data

Conclusion: a step-by-step approach to ML implementation and the evaluation of vendor solutions

Target Audience

This workshop is aimed at R&D and QC engineers as well as managers and group leaders wanting to understand the key aspects of machine learning, both in a context of decision-making of AI-based vendor products or for in-home implementation of custom ML solutions.