A novel activation function based on DNA enzyme-free hybridization reaction and its implementation on nonlinear...

Area of Science:

• Biotechnology
• Computer Science
• Artificial Intelligence

Background:

• Traditional silicon-based computing faces limitations in the post-Moore's Law era.
• DNA computing offers powerful parallel processing and data storage capabilities.
• Activation functions are crucial for AI but difficult to implement in DNA circuits.

Purpose of the Study:

• To propose a novel, easily implementable activation function for DNA computing.
• To enable DNA circuits to perform complex AI functions, including nonlinear fitting and prediction.
• To integrate DNA computing with artificial intelligence for advanced applications.

Main Methods:

• Developed a new activation function based on DNA enzyme-free hybridization reactions and DNA molecule displacement kinetics.
• Provided mathematical and kinetic analyses of the proposed activation function.
• Nested the activation function into a nonlinear neural network for DNA computing.

Main Results:

• The proposed activation function is readily implementable using enzyme-free DNA reactions.
• The function exhibits favorable nesting and cascading properties for DNA circuit construction.
• The DNA-based neural network successfully fitted and predicted a nonlinear function.

Conclusions:

• This work presents a significant advancement in DNA computing for artificial intelligence applications.
• The novel activation function overcomes key limitations in implementing AI within DNA circuits.
• The developed system demonstrates the potential of DNA computing for complex nonlinear data processing.