PROJECT TITLE :

Online ADMM-based Extreme Learning Machine for Sparse Supervised Learning

ABSTRACT:

In the field of Machine Learning, sparse learning is a useful strategy for selecting features and avoiding overfitting. An online sparse supervised learning of extreme learning machine (ELM) technique based on alternate direction method of multipliers (ADMM) is developed for real-world situations with Online Learning needs in Neural Networks, dubbed OAL1-ELM. An l 1 -regularization penalty is introduced to the loss function in OAL1-ELM to generate a sparse solution and improve generalization ability. This convex combinatorial loss function is solved in a distributed manner using ADMM. In addition, to reduce computing complexity and achieve online learning, an upgraded ADMM is implemented. The proposed method is capable of learning data one by one or in batches. The suggested method's efficiency and optimality are demonstrated through a convergence analysis for the fixed point of the solution. The suggested method can find a sparse solution and has high generalization performance in a wide range of regression tasks, multiclass classification tasks, and a real-world industrial project, according to the experimental results.