StreaMRAK a streaming multi-resolution adaptive kernel algorithm

Kernel ridge regression (KRR) is a popular scheme for non-linear non-parametric learning. However, existing implementations of KRR require that all the data stored in main memory, which severely limits use contexts where size far exceeds memory size. Such applications are increasingly common mining, bioinformatics, and control. A powerful paradigm computing on sets too large streaming model computation , we process one sample at time, discarding each before moving to next one. In this paper, propose StreaMRAK - version KRR. improves schemes by dividing problem into several levels resolution, allows continual refinement predictions. The algorithm reduces requirement continuously efficiently integrating new samples training model. With novel sub-sampling scheme, computational complexities creating sketch original data, density adapted bandwidth kernel local dimensionality data. We present showcase study two synthetic problems prediction trajectory double pendulum. results show proposed fast accurate.