Source-linked AI summary
Memristive model of amoeba's learning
Yuriy V. Pershin, Steven La Fontaine, Massimiliano Di Ventra
TL;DR
The paper addresses how Physarum memorizes periodic environmental changes and anticipates future stimuli. It models this behavior with an LC circuit and memristor driven by voltage pulses, proposing that state-dependent gel-sol channel changes provide a possible biological analogue. The circuit reproduces stimulus-timed slowdowns, trained anticipation after a fourth pulse, and learning concentrated near the LC timescale.
Problem
Physarum shows sequence memory and anticipatory behavior, while oscillator-based explanations do not fully explain its memory response or other history-dependent physiological changes.
Method
The paper maps Physarum’s movement and memory mechanisms onto a passive RLC circuit with a voltage-controlled memristor driven by environmental-mimicking voltage pulses.
Results
The circuit reproduces learned, stimulus-timed slowdowns, trained anticipation after a fourth pulse, and significant memristor changes for pulse intervals of 8s< τ <10s.
Takeaways & Limitations
A passive memristive circuit can model Physarum’s pattern recognition and event prediction while providing a dynamic picture of a possible biological memory mechanism.
Abstract
from arXiv · showhide
Recently, it was shown that the amoeba-like cell {\it Physarum polycephalum} when exposed to a pattern of periodic environmental changes learns and adapts its behavior in anticipation of the next stimulus to come. Here we show that such behavior can be mapped into the response of a simple electronic circuit consisting of an $LC$ contour and a memory-resistor (a memristor) to a train of voltage pulses that mimic environment changes. We also identify a possible biological origin of the memristive behavior in the cell. These biological memory features are likely to occur in other unicellular as well as multicellular organisms, albeit in different forms. Therefore, the above memristive circuit model, which has learning properties, is useful to better understand the origins of primitive intelligence.
I. INTRODUCTION
Physarum can memorize periodic environmental sequences and anticipate later stimuli, but existing oscillator-based explanations do not fully account for this memory. The paper proposes a memristive circuit analogy linking state-dependent biological changes to learning behavior.
- Physarum can memorize periodic temperature and humidity changes, then slow at the times when subsequent stimuli are expected.The learned pattern dissipates over time, but a single stimulus can reactivate the oscillations within a certain time frame.
- The authors restrict “learning” to the primitive sequence-memory behavior observed in recent Physarum experiments.They distinguish it from classical conditioning and associative memory observed in more developed animals.
- An oscillator-based model suggests that stimulus frequencies excite internal biological oscillators, but it does not fully explain the amoeba’s memory response.It also omits possible microscopic physiological changes that depend on the system’s prior states.
- State-dependent physiological changes are proposed as a more likely source of memory effects than oscillator excitation alone.The authors identify a possible mechanism while acknowledging that the actual realization of memory in amoebas remains unknown.
- The proposed biological mechanism involves a thixotropic gel-sol medium whose pressure-dependent viscosity can form low-viscosity channels, analogous to memristive resistance changes.These channels may alter sol flow and preserve a history-dependent state.
- The paper maps Physarum’s movement mechanisms onto a passive circuit containing a resistor, inductor, capacitor, and memristor.The model treats environmental conditions as external voltage and the memristor as the element summarizing memory mechanisms.
II. MEMRISTIVE CIRCUIT
The paper maps Physarum’s adaptive behavior onto an electronic circuit combining an LC contour with a memristor whose state stores stimulus history. The model links circuit elements to biological processes and simulates responses to applied voltage-pulse sequences.
- Circuit–biology mapping: The circuit uses a resistor, capacitor, inductor, and memristor to model biological processes involved in Physarum’s movement and learning.The analogy is explicitly presented as a simplification of the biological learning process.
- Simulation setup: Simulations compare arbitrary and learning pulse sequences, and model spontaneous slowdown responses after periodic three-pulse training.The figures use applied pulse sequences to represent changing environmental conditions.
- Memristive mechanism: The memristor stores information about past pulses and controls oscillations in the LC contour through its internal resistance state.Its resistance changes between limiting values M1 and M2, with M1 < M2.
- Memristive mechanism: The memristor-state equation constrains resistance changes between M1 and M2 according to the capacitor voltage and a step-function activation rule.The capacitor voltage equals the memristor voltage in the circuit model.
- Memristive mechanism: Learning is modeled as faster when |VC| > VT and slower when |VC| < VT, with VC driving the memristor-state change.The threshold voltage VT determines the activation-dependent learning rate.
- Simulation setup: The circuit response is obtained by numerically solving the circuit equations with initial conditions near steady state under different applied voltage signals.The equations describe voltage balance and current conservation for the circuit elements.
III. RESULTS AND DISCUSSION
The circuit reproduces amoeba-like anticipation: periodic pulses near the LC resonance strengthen memristor learning and produce subsequent slowdown events, whereas irregular pulses produce weaker responses. The model links this behavior to state-dependent signal conservation and dynamic vein formation.
- Circuit response: Periodic pulses near the LC contour’s resonant frequency drive a larger memristance change and sustain oscillations longer by reducing damping.The capacitor voltage grows across resonant pulses until it exceeds the memristor threshold, switching the memristor toward higher resistance.
- Circuit response: Three regular training pulses reproduce spontaneous slowdown at the expected subsequent pulse times, matching the reported amoeba behavior.The same circuit response is substantially smaller after three irregular training pulses.
- Circuit response: A fourth pulse elicits several well-defined post-pulse slowdown events after periodic training but no significant anticipated slowdown after non-periodic training.This models the reported SPSD distinction between previously trained and untrained organisms.
- Resonance dependence: Learning occurs for pulse intervals 8s< τ <10s, close to the LC contour timescale, and increasing β makes this learning interval more sharply defined.The plotted quantity is the memristor state immediately after the third learning pulse.
- Resonance dependence: Significant changes in M occur when the applied frequency approaches resonance because capacitor-voltage oscillations then exceed the memristor threshold.Away from resonance, LC oscillation amplitude decreases, limiting memristor-state changes.
- Biological interpretation: The biological analogy maps rhythmic actin-fiber contractions and pressure-dependent vein remodeling onto LC oscillations and memristive state change.The model treats evolving vein formation and degradation as a simplified analogue of memory-bearing circuit dynamics.
IV. CONCLUSIONS
The proposed memristive circuit models Physarum’s pattern recognition and event prediction while providing a dynamic account of memory in the organism. Because it uses passive elements and can be extended to multiple learning elements, the model may support circuit-learning applications and studies of primitive intelligence.
- The electronic circuit closely simulates Physarum’s experimentally observed ability to recognize patterns and predict events.
- Its memristive element represents biological memory mechanisms in the protoplasm, producing sol flow dependent on the system’s history and state.
- The model provides a dynamic picture of memory in this unicellular organism, including memory of the stimulus period.
- Because the circuit uses only passive elements and is laboratory-realizable, it may support electronic applications requiring circuit learning or pattern recognition.
- Extensions with multiple learning elements may find applications in neural networks and help investigate primitive intelligence and adaptive behavior.