Reservoir Computing With Dynamic Reservoir using Cascaded DNA Memristors

Area of Science:

• Molecular electronics
• Biomolecular computing
• Nanotechnology

Background:

• Traditional molecular and DNA memristors use two output variables, preventing cascading due to mismatched input/output sizes.
• Cascading is crucial for building complex computational systems like reservoir computing (RC) models.

Purpose of the Study:

• To propose molecular and DNA memristors with a single output variable for state definition.
• To enable cascading of these memristors for enhanced reservoir computing applications.
• To reduce the complexity and resource requirements of biomolecular computing systems.

Main Methods:

• Redefining the state variable in molecular and DNA memristors to a single output.
• Cascading the proposed memristors to form dynamic reservoirs for RC.
• Developing a DNA-based RC system for seizure detection from iEEG data.
• Utilizing the DNA RC system for time-series prediction tasks.

Main Results:

• The proposed single-variable state definition allows for successful cascading of molecular and DNA memristors.
• Cascaded memristors exhibit retained memristive behavior and enable dynamic reservoir synthesis.
• A DNA RC system effectively detected seizures from iEEG data.
• The DNA RC system accurately performed time-series prediction, reducing DNA strand displacement reactions by 3-5 times.

Conclusions:

• The novel memristor design facilitates cascading, leading to more efficient and scalable biomolecular computing architectures.
• Dynamic reservoirs built with these memristors offer significant advantages over static ones.
• The demonstrated DNA RC systems show promise for real-world applications in biosignal processing and predictive modeling.