Meminductor Achieves Novel Neuromorphic Computing, Replicating Amoeba Behaviour Via Charge-Dependent Inductance
Summarize this article with:
The pursuit of brain-inspired computing has led researchers to explore new electronic components that mimic biological processes, and now Frank Zhigang Wang from the University of Kent, along with colleagues, presents a significant advance beyond the memristor. This team demonstrates the functionality of the meminductor, an inductor whose inductance changes based on the history of current flowing through it, effectively giving it a memory. Unlike traditional resistors with memory, the meminductor’s behaviour, tied to the time constant of electrical circuits, unlocks unique possibilities for applications in deep learning and recreating complex biological behaviours. The researchers experimentally validated this new component by successfully reproducing the memorizing, timing and anticipating mechanisms observed in amoebae, establishing a theoretically sound and practically viable alternative to existing neuromorphic computing approaches. An inductor with a magnetic core functions as a meminductor, characterized by its inductance L(q) changing with the electric charge, q, passing through the coil. The magnetic core retains a history of current flow, effectively giving the inductor a memory function, and offering a unique capability distinct from that of a memristor in neuromorphic computing, deep learning, and brain-inspired systems. Researchers demonstrated this concept by successfully reproducing the observed biological behaviour of amoebae, specifically relating to memorizing and timing events.
Memristors Emulate Biological Neural Processes This research details advancements in computing architectures, focusing on memristors and their potential to mimic biological neural processes. These nanoscale devices change their resistance based on current flow, mimicking synaptic plasticity in biological neurons, and can create artificial neurons and synapses forming the basis of new neural networks. Their ability to remember past electrical activity is crucial for implementing learning and adaptation. A significant focus is on delayed switching in memristors, which improves performance and functionality in neural network behaviour. The work draws parallels between memristor-based systems and the fundamental principles of biological intelligence, including primitive intelligence observed in simple organisms like amoebas, suggesting that understanding primitive intelligence can inform the design of more efficient artificial intelligence systems. The research addresses limitations of traditional von Neumann architectures, proposing that memristor-based systems overcome these limitations by integrating memory and processing into a single device, leading to improvements in energy efficiency, speed, and scalability. Potential applications include creating more powerful AI systems, building brain-inspired computers, developing pattern recognition systems, creating robots with advanced cognitive abilities, and developing novel computing platforms for healthcare. Dynamic Inductance and Meminductor Realization Scientists have invented a new circuit element, the meminductor, which expands beyond the capabilities of traditional memristors and opens possibilities for novel computing architectures. This meminductor is realized using a coil with a magnetic core, where the inductance, L, dynamically changes as a function of the electric charge, q, passing through the coil. The core’s magnetization remembers the history of current flow, effectively giving the inductor a memory function.
The team demonstrated that the meminductor’s behaviour is governed by a relationship where the rate of change of magnetic flux is directly proportional to the applied voltage, accurately described by integrating the current over time. Measurements confirm that the inductance, L(q), decreases as charge accumulates, while maintaining essential characteristics of an ideal circuit element, including non-linearity, continuous differentiability, and monotonic increase. To validate the concept, researchers used the meminductor to simulate the behaviour of amoebae, specifically their ability to memorize, time events, and anticipate future stimuli. A neuromorphic circuit, incorporating the meminductor, was designed to mimic an amoeba’s response to temperature changes. Simulations and experiments reveal that the circuit’s resonance frequency, determined by the dynamic inductance L(q), scans a range of frequencies, triggering a resonance when matching the frequency of an external stimulus, mirroring the amoeba’s response to temperature drops. Remarkably, the circuit accurately reproduces the amoeba’s complex behaviour, including a sustained slow-down even when the stimulus does not occur at anticipated times. Experiments using a hardware emulator confirm the simulation results, demonstrating the meminductor’s ability to mimic the amoeba’s memorizing, timing, and anticipating mechanisms. Meminductor Realization and Circuit Implications This research demonstrates the creation and characterization of a meminductor, an inductor whose inductance varies with the amount of electric charge passing through it.
The team established that a coil incorporating a magnetic core functions as a meminductor because the core’s magnetization retains a history of current flow, effectively altering the inductance. This contrasts with traditional inductors which maintain a constant inductance regardless of current history. The significance of this achievement lies in expanding the possibilities for novel computing architectures, particularly in areas like neuromorphic computing and deep learning. Unlike memristors, which modulate resistance, meminductors affect the time constant of circuits, offering a different mechanism for processing information and potentially enabling more complex and efficient systems. Researchers successfully reproduced the memorizing, timing, and anticipating behaviours observed in amoebae using a meminductor-based circuit. The authors acknowledge that further investigation is needed to fully explore the capabilities of this meminductor and its integration into larger-scale computing systems, suggesting that this architecture holds promise for understanding the cellular origins of primitive intelligence and developing innovative approaches to computation. 👉 More information 🗞 Beyond Memristor: Neuromorphic Computing Using Meminductor 🧠 ArXiv: https://arxiv.org/abs/2512.11002 Tags: