Inverse-Free Extreme Learning Machine With Optimal Information Updating

Area of Science:

• Machine Learning
• Artificial Intelligence
• Computational Science

Background:

• Extreme Learning Machines (ELM) are effective for various machine learning tasks.
• The standard ELM algorithm's reliance on matrix inversion presents computational challenges, limiting its scalability.
• Efficient algorithms are needed to broaden ELM's applicability.

Purpose of the Study:

• To develop an inverse-free extreme learning machine (ELM) algorithm.
• To address the computational limitations of traditional ELM methods.
• To enable incremental increases in hidden nodes and optimal weight updates.

Main Methods:

• An incremental approach to increasing hidden nodes.
• Progressive and optimal updating of connection weights.
• Theoretical analysis to demonstrate error reduction and optimality.

Main Results:

• The proposed inverse-free ELM algorithm demonstrates monotonic decrease in training error.
• Each updating step in the algorithm is proven to be optimal.
• Extensive numerical experiments validate the algorithm's effectiveness and accuracy.

Conclusions:

• The inverse-free ELM offers a computationally efficient alternative to standard ELM.
• The incremental and optimal weight updating strategy enhances performance.
• This method significantly expands the practical applications of ELM.