Delogne, Rémi
[UCL]
Jacques, Laurent
[UCL]
Random embedding techniques, such as random Fourier features, are widely used to sketch initial data to a new, kernelised feature space. In this work, we leverage a specific property of random rank-one projection operators, the sign product embedding, to approximate a quadratic polynomial kernel using the scalar product of a pair asymmetric vector embeddings, with one taking only binary values. We demonstrate empirically that the approximated kernel compares favourably to the initial one on toy binary classification examples.
Bibliographic reference |
Delogne, Rémi ; Jacques, Laurent. Quadratic polynomial kernel approximation with asymmetric embeddings.DEEPK24 (Leuven, du 07/03/2024 au 08/03/2024). In: International Workshop on Deep Learning and Kernel Machines (2024), |
Permanent URL |
http://hdl.handle.net/2078.1/285963 |