Probability Seminars
Rough Paths and more scalable data science
With Terry Lyons (Oxford)
Rough Paths and more scalable data science
The basic principle of Rough Path Theory, is that one can understand multi modal and oscillatory streams of data by considering their impact on non-linear controlled dynamical systems.
The approach leads to the local description of these complex streams through signatures; connecting data science with tensor algebra.
These high order signatures of the data are expensive to compute but can be re-used a huge number of times in the training of a neural net without the need to re-compute the signature from the data.
- Speaker: Terry Lyons (Oxford)
- Monday 05 September 2022, 11:30–12:30
- Venue: MR9, Centre for Mathematical Sciences.
- Series: Probability; organiser: Jason Miller.