Accelerated Elastic Network Regression via Random Projection

In the large-scale data prediction, in order to better describe the overall characteristics of the data matrix and ensure the model accuracy, an elastic network ENLF model using L
₁
and L
₂
limiting objective functions is usually constructed. ENLF model improves the performance of the model, but it has low time effect and high computational complexity for large-scale data matrix applications. The k-low rank approximation dimension reduction of large-scale data is carried out by introducing stochastic projection algorithm, and the model is solved by using accelerated proximal gradient descent algorithm and coordinate descent algorithm. Theoretical analysis and experimental results show that the correlation algorithm based on random projection has better convergence speed, time efficiency and lower computational complexity than the traditional algorithm in solving the ENLF model of large scale data, and the accuracy of the model can be guaranteed.