Course Area
Availability
Available for 365 days after enrollment
Approximate Course Run Time
3 hours
Continuing Education Units
0.3
Course Fee
USD $500.00
E-learning content is available on demand
Course Details
Assessing robustness of AI systems is a crucial part of their development and deployment. Here at BSI, we have developed a course that explores in depth the importance and impact of implementing robust deep learning systems, following the Standard ISO/IEC 24029-1:2021. We deep dive into the core of the standard, exploring all its technical parts and the corresponding actionable contents.
On-demand - training that’s even more flexible
BSI’s on-demand courses are market-leading and available 24/7. Developed by top subject matter experts, they contain the same high-quality content you will find in our tutor-led training, but with the added benefit of being able to learn at your own pace and at any time.
Course aim
The aim of this course is to give you the technical tools to ensure that your deep learning models are robust, i.e., they perform well all circumstances, even in unforeseen and unusual situations.
We aim to give you knowledge of the main technical concepts of ISO/IEC 24029-1:2021, by explaining the different methods (statistical, formal, empirical) through which robustness can be assessed, and some more advanced concepts, such as data perturbation and abstract interpretation.
The course helps reinforce the importance of assessing robustness and understanding the impact that robustness issues can have on the development and deployment of deep learning systems
We aim to give you knowledge of the different contexts in which robustness concerns can arise, by providing examples of robustness concerns and give possible strategies for how to detect (and possibly address) such issues
How will I benefit?
This course will help you:
- Build a toolset of methods that can help identifying robustness issues
- Understand the different types of data perturbations, and their use in the creation robustness test data sets
- Design workflows to detect and address robustness concerns
- Take steps to ensure that the assessment of robustness is part of the development and deployment of AI systems involving neural networks
What will I learn?
Upon completion of this course, you will be able to:
- Apply the methods provided by the standard to detect robustness issues arising in the development and deployment of a deep learning system
- Describe the main principles behind data perturbation and abstract interpretation
- Construct protocols to assess robustness of a neural network
Modules
- Module 1 - statistical methods of robustness
- Module 2 - formal methods for robustness
- Module 3 - empirical methods for robustness
- Module 4 - annexes of the standard
- module 5 - review and summary