You are all cordially invited to the next AMLab colloquium on Tuesday, January 12 at 16:00 in C3.163, where Albert Huizing, principle scientist at TNO, will give a talk titled “Applying the art of deep learning to radar”.
Abstract: Deep learning techniques are currently successfully applied to object recognition in images, face recognition and speech recognition by major companies such as Google, Microsoft, Facebook, Apple, and Baidu. The application of deep learning to radar is at a more embryonic level than optical and acoustical sensors, partly because of the unknown benefits of radar in machine learning communities and partly because of the lack of large radar signal databases for training deep neural networks. The similarity of radar and optical images and radar and acoustical spectrograms, and the predicted growth of radar in commercial applications such as autonomous driving and gesture recognition, indicate that there is a need for investigating the application of deep learning in radar.
This talk will give a brief introduction of the principles of radar and provide a live demonstration of a small radar system. An overview of radar applications will show the benefits of radar when compared with other sensors such as cameras and microphones. In a case study, a model-based and a feature-based approach to human motion analysis with radar will be presented. The potential of deep learning for human motion analysis with radar will be discussed and concepts for fusing the signals of radars and cameras with multimodal deep neural networks will be explored.
You are all cordially invited to the next AMLab colloquium on Tuesday, December 15 at 16:00 in C3.163, where John Ashley Burgoyne from the music cognition group here at the UvA will give a talk titled “‘Big data’, music-cognition style”.
Abstract: Machine learning techniques have been fundamental in the development of music information retrieval (MIR), but there has been much more resistance to applications in music cognition (Aucouturier & Bigand, 2013). As more cognition experiments collect relatively large amounts data, however – Internet-based experiments, for example, or recordings of live environments – interest in machine learning is growing.
This talk will introduce two ongoing projects at the University of Amsterdam’s Music Cognition Group that could benefit from machine-learning approaches. The first, Hooked on Music, is an online music game, now played more than 3 million times, that seeks to identify which musical characteristics may be responsible for long-term musical memories. The modelling challenge is balancing predictive accuracy with interpretability from a musical perspective. The second is a series of overhead video recordings of silent discos, popular dance events during which music is streamed to participants on multi-channel headphones. Here the modelling challenge is in the pre-processing: we would like to test various hypotheses about music and social behaviour, but this is only possible if we can get reasonably accurate location tracking of the LED lights on each set of silent-disco headphones.
In both cases, I will discuss our modelling approaches and results to date, where those models might be improved using machine learning approaches, and how it might be most interesting to collaborate.
You are all cordially invited to the next AMLab colloquium on Tuesday, December 1 at 16:00 in C3.163, where Sara Magliacane will give a talk titled “Probabilistic logical causal inference”.
Abstract: Discovering causal relations from data represents the core of the scientific method. In most cases the causal relations are recovered from experimental data in which the variable of interest is perturbed, but seminal work from Spirtes and Pearl demonstrates that, under certain assumptions, it is already possible to exclude several implausible causal models of the data by using only observational data. Constraint-based causal discovery methods use statistical (in)dependences from the data to express constraints over all the possible causal models. One of the most promising formulations of this problem is in logic, which allows for quick prototyping, combination of algorithms and an easy integration of complex background knowledge. On the other side, a purely logic approach cannot handle noise in the (in)dependence test results, making the case for the use of probabilistic logics.
In this talk, I will present two algorithms for probabilistic logical causal inference that we have been developing with Tom Claassen from RU Nijmegen. Compared to other existing methods, our algorithms are more scalable and simpler to encode, while preserving a comparable accuracy in the prediction of indirect causal and acausal relationships.