Geopolymer Memristors in Computing

The memristor-based RC system typically consists of two key components. The first is the reservoir, composed of an array of memristors fabricated using nanoscale manufacturing techniques. These memristors are integrated into the system and interfaced with input sources, such as images of handwritten digits or other physical signals. As temporal input is applied, each memristor responds dynamically, and their combined states—referred to as the reservoir state—enable complex signal processing.

This reservoir performs nonlinear transformations of the input data in real time, mapping it into a high-dimensional space. The second component, the readout layer, is a simple and easily manufacturable circuit that applies fixed random weights to the reservoir output. Since only this layer requires training (e.g., using linear or logistic regression), it simplifies hardware implementation and shortens development cycles.

Overall, the architecture's reliance on memristors and minimal training overhead makes it well-suited for scalable, energy-efficient, and cost-effective manufacturing of AI hardware for edge computing and other real-time applications..