#27: MLDublin meets ADAPT @ DogPatch Labs

Machine learning models such as deep Neural Networks, Random Forests andSupport Vector Machines provide effective solutions for many applications.Nevertheless, these approaches are currently mostly used as a black-box -the user cannot understand their decision process.This problem gets worse when the relevant input data already has a complicated structure.For example, the input patterns may be arranged in a particular order, such as a sequence,tree or graph, and the combination of possible patterns creates a very complex feature spaceand typically leads to a large and complex prediction model.For successful machine learning adoption, we need explainable machine learning models:new learning algorithms that are effective, and whose outcome is easy to understand by the end user.This talk briefly discusses some of our work on developing accurate machine learningalgorithms for sequences and time series, while keeping the interpretability of the model as a coredesign constraint.

AI Podcaster Mark Kelly will share his experience of how companies are applying AI in Ireland and will also talk about the AI Awards happening later this year.