The talk will address: building models simple enough to be explainable, but complex enough to explain the data; adaptable to new data arriving; estimating the confidence in predictions; generating the model space.
A technique is introduced which builds the one-dimensional model space consisting of one basis function. More basis functions are added, if informed so by the data. Basis functions can be deleted if a more suitable one is found. The possible choices are provided from a dictionary. A geometric explanation is given, how the goodness of basis functions for the model is assessed. This technique can update the model, if new data arrives, which is suitable when a sparse model is needed for example due to limited up-link and/or down-link bandwidth. It can also assess the confidence in predictions using the expected change in likelihood. The talk concludes outlining work in progress on the generation of basis functions.
Bio: Anita Faul came to Cambridge after studying two years in Germany. She studied Part II & Part III Maths at Churchill College, Cambridge. This was followed by a PhD on the Faul-Powell Algorithm for Radial Basis Function Interpolation under the supervision of Prof Mike Powell. This collaboration resulted in an Erdős number of 4 (Powell - Beatson - Chui - Erdős). Current projects are on ML techniques. In teaching she enjoys to bring out the underlying, connecting principles of algorithms which is the emphasis of her book on Numerical Analysis.
*Sponsors*
Man AHL: At Man AHL, we mix machine learning, computer science and engineering with terabytes of data to invest billions of dollars every day.
[ Ссылка ] : Machines that Read - Intelligent data extraction from corporate and financial documents.
