Program Radboud AI Event
11:45 Walk in with sandwiches
12:15 Introduction AI prof. dr. Marcel van Gerven
13:00 Advances in deep learning dr. Jonas Teuwen On of the most active research directions in deep learning nowadays is in generative modelling where one tries to model a probability distribution given samples drawn from that distribution. As these models intend to capture the underlying distribution, they are an important candidate for many unsupervised learning problems.
Therefore, over the past few years, many different generative models have been proposed. In this talk we will focus on Normalizing Flows, which are a class of generative models which produce tractable distributions where both sampling and density evaluation are computationally efficient and exact. We will also discuss their connection to invertible neural networks which have independently attracted much interest due their favourable memory requirements and their connections to differential equations.
13:45 Probabilistic Machine Learning dr. Luca Ambrogioni The masterclass will cover the basic principle behind Bayesian deep learning and variational inference. After an introduction into stochastic variational inference, variational autoencoders and the reparameterization trick we will explain more advanced advanced topics such as Wasserstein variational inference and normalizing flows.
14:30 Coffee break
14:45 Large-scale Federated Search prof. dr. Djoerd Hiemstra In federated search, independent search engines owned and operated by different stakeholders cooperate to collectively provide a search service. In the masterclass: • several scenarios will be discussed with minimal cooperation between search engines • various use cases will be showed such as enterprise search • machine learning approaches will be discussed based on building language models from the results obtained from search engines, as well as future work that used externally trained language models and transfer learning as an effective approach to federated search.
15:30 Speech Technology dr. Henk van den Heuvel Automatic speech recognition is a powerful means to get to grips with your audio and video data. Applications are manyfold nowadays. An overview will be presented of inspiring and surprising examples, showing thestrength of the technology but also its weaknesses. Some of these weaknesses have been tackled and will be shown in recent applications.
16:15 Wrap-up prof. dr. Marcel van Gerven
16:45 Network drinks
Who should attend and why?
AI practitioners such as software engineers, data-scientists, data analysts and researchers.
You will:
• gain insight into the newest developments of AI
• gain the newest practical and scientific insights into Machine Learning
• expand your network
• become part of the Radboud AI network