Laboratory of Signal Processing
Tampere University of Technology
Computational Analysis of Sound Events in Realistic Multisource EnvironmentsRealistic everyday soundscapes often consists of multiple concurrently present sound sources, containing important information for many applications. Recent development in machine learning, especially deep learning has allowed the development of methods to automatically recognize sounds even in these challenging multisource environments. This talk will give an overview of the recent developments in the field. We will start by first discussing the typology of various audio analysis tasks, including classification, detection, and tagging, of both higher-level sound scene categories and also individual sound events. We will then present specific methods based on deep neural networks, which use multilabel classification for efficient multi-source recognition. We will present a convolutional recurrent neural network based approach that has successfully been used in many scene analysis problems. We present how such a network allows simultaneously learning relevant acoustic features for recognition and modeling longer temporal context. The talk will also summarize findings from the recent public DCASE evaluation campaigns.
Scientific Computing & Imaging Institute
Huntsman Cancer Institute
University of Utah, USA
Comparative Spectral Decompositions for Personalized Cancer Diagnostics and PrognosticsI will describe the development of novel, multi-tensor generalizations of the singular value decomposition, and their use in the comparisons of brain, lung, ovarian, and uterine cancer and normal genomes, to uncover patterns of DNA copy-number alterations that predict survival and response to treatment, statistically better than, and independent of, the best indicators in clinical use and existing laboratory tests. Recurring alterations have been recognized as a hallmark of cancer for over a century, and observed in these cancers’ genomes for decades; however, copy-number subtypes predictive of patients’ outcomes were not identified before. The data had been publicly available, but the patterns remained unknown until the data were modeled by using the multi-tensor decompositions, illustrating the universal ability of these decompositions – generalizations of the frameworks that underlie the theoretical description of the physical world – to find what other methods miss.
Brain Science Institue
Tensor Networks and their Potential Applications in Dimensionality Reduction and Blind Signal ProcessingIn this talk we discuss briefly tensor networks which provide a natural sparse and distributed representation for large scale data, and address both established and emerging methodologies for tensor-based decomposition and optimization. Our particular focus will on low-rank tensor network representations, which allow for huge data tensors to be approximated (compressed) by interconnected low-order core tensors. The usefulness of this concept is illustrated over a number of applied areas, including multiway analysis, multilinear ICA/BSS, deep learning, generalized regression, tensor canonical correlation analysis and higher order partial least squares. Special emphasis will be given to the links between tensor networks and deep learning and abilities of some specific tensor networks that significantly compress both the fully connected layers and the convolutional layers of deep neural networks.
- Cichocki, A., Lee, N., Oseledets, I., Phan, A. H., Zhao, Q., & Mandic, D. P. (2016). Tensor Networks for Dimensionality Reduction and Large-Scale Optimization: Part 1 Low-Rank Tensor Decompositions. Foundations and Trends® in Machine Learning, 9(4-5), 249-429. https://arxiv.org/abs/1609.00893
- Cichocki, A., Phan, A. H., Zhao, Q., Lee, N., Oseledets, I., Sugiyama, M., & Mandic, D. P. (2017). Tensor Networks for Dimensionality Deduction and Large-Scale Optimization: Part 2 Applications and Future Perspectives. Foundations and Trends® in Machine Learning, 9(6), 431-673. https://arxiv.org/abs/1708.09165
- Cichocki, A., Mandic, D., De Lathauwer, L., Zhou, G., Zhao, Q., Caiafa, C., & Phan, H. A. (2015). Tensor Decompositions for Signal Processing Applications: From Two-Way to Multiway Component Analysis. IEEE Signal Processing Magazine, 32(2), 145-163.