How Hopfield Networks Use Resistors to Mimic Brain-Like Memory

A foundational patent describing an electronic circuit that uses a grid of resistors to perform computations, effectively creating an artificial neural network that can store and recall patterns.

What it covers

This patent describes a hardware architecture where multiple amplifiers are connected in a dense matrix using resistors. Each resistor acts as a weight, determining how much the output of one amplifier influences the input of another. By setting these resistance values, the network can be programmed to store specific patterns or solve optimization problems. When the system is given a partial or noisy input, the interconnected resistors cause the circuit to settle into a stable state, effectively reconstructing the full stored pattern or finding a solution to a complex problem.

What it doesn't cover

• —Does not cover software-based neural network simulations running on standard CPUs.
• —Does not cover digital logic gates or traditional von Neumann computing architectures.
• —Does not cover learning algorithms that automatically adjust resistor values during operation (this patent focuses on fixed, programmed resistances).
• —Does not cover non-electronic implementations, such as optical or biological neural models.

The clever bit

The innovation lies in using the physical laws of electricity—specifically Kirchhoff's circuit laws—to perform computation. Instead of calculating results step-by-step, the network 'finds' the solution by naturally evolving toward a state of lowest energy.

Why it matters

This is a seminal document in the history of artificial intelligence. John Hopfield's work provided a physical, electronic blueprint for associative memory, bridging the gap between theoretical neuroscience and practical circuit design. It proved that simple, collective interactions between basic components could produce complex, brain-like computational behavior.

Real-world examples

• 1.Early hardware-based associative memory systems
• 2.Neuromorphic computing research platforms
• 3.Analog neural network circuit prototypes