This report explores using the Foward-Foward (FF) algorithm to train neural networks. Unlike traditional training methods that use a forward and backward pass, the FF algorithm uses two forward passes with opposite objectives. The findings of this report lay the groundwork for feature work on the FF algorithm and propose a new method to control the influence individual layers have on the network's performance.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Leonard, Jared

Publisher

• Worcester Polytechnic Institute

Identifier

• E-project-121423-170542
• 115204

Palavra-chave

• Layer-wise Learning
• Forward-Forward Algorithm
• Neural Network Training
• Goodness Score Maximization
• MNIST Dataset Classification
• Gradient Descent Alternatives

Advisor

• Paffenroth, Randy Clinton

Year

• 2023

Date created

• 2023-12-14

Resource type

• Major Qualifying Project

Major

• Data Science

Source

• E-project-121423-170542

Rights statement

• In Copyright

Última modificação

• 2024-02-07