Prof. Peter Tiňo: Learning from Temporal Data Using State Space Models
Speaker: Prof. Peter Tiňo (University of Birmingham)
Title: Learning from Temporal Data Using State Space Models
State space modelling is a popular approach to processing data with temporal dependencies. There is a huge variety of state space models, but they all share the same common underlying modelling principle: Collapse all the 'relevant' information from the past into an ‘abstract’ information processing state, which can be updated recursively as more and more data arrives. Outputs of such models are then determined based on the state information (representing the past) and the current data.
In machine learning, state space models take on many forms in various contexts, depending on e.g.
- the nature of the state space and its cardinality (e.g. finite number of states, discrete state space, continuous state space)
- whether temporal mapping of inputs to outputs, or modelling of input time series is considered (supervised vs. unsupervised learning)
- the model formulation framework - e.g. probabilistic vs. non-probabilistic
Some well-known model classes include recurrent neural networks (including e.g. LSTM and Echo State Networks), finite state machines, hidden Markov models, (extended) Kalman filters etc. I will present a unified view of such systems as non-autonomous input-driven dynamical systems and will outline general issues (such as the difficulty of latching important information from the past, or non-observable states) that are common to different formulations of state space models.
Peter Tiňo is a Professor of Complex and Adaptive Systems at the School of Computer Science at the University of Birmingham. He is the author of over 160 research articles in the areas of dynamical systems, machine learning, natural computation and fractal geometry. Peter has been awarded three outstanding Journal Paper of the Year awards and the Head of School's Excellence in Teaching Award.
Professor Tiňo is a co-supervisor of ECOLE, an Innovative Training Network (ITN) for early stage researchers (ESRs) funded by the EU’s Horizon 2020 research and innovation program under grant agreement No.766186. It is based on novel synergies between nature inspired optimisation and machine learning. The training programme will be targeted at the automotive industry and ESRs employed on the program will be provided with the transferable skills necessary for thriving careers in emerging and rapidly developing industrial areas.