Chain-block algorithm to RVM on large scale problems
GangLi, Shu-BaoXing, Huifeng Xue · 2009
RVM enables sparse classification and regression functions to be obtained by linearly-weighting a small number of fixed basis functions from a large dictionary of potential candidates.TOA on RVM has O(M3) time and O(M2) space complexity, where M is the training set size. It is thus computationally infeasible on very large data sets. We propose CBA . it decomposed large datasets to subdata blocks by sampled homogeneously and getted solution by chain iteration taking TOA as basis algorithm. Regression experiments with synthetical large sbenchmark data set demonstrates CBA yielded state-of-the-art performance: its time complexity is linear in M and space complexity is independent of M, keeping high accuracy and sparsity at the same time. Document shows that CBA is also much better than TFA on time complexity and sparsity.