arXiv subjects

Miguel Ruiz-Garcia

Publications and source records attributed to Miguel Ruiz-Garcia.

17 recordsLinked to original sources

arXiv2025

Collective Behavior and Memory States in Flow Networks with Tunable Bistability

Multistability-induced hysteresis has been widely studied in mechanical systems, but such behavior has proven more difficult to reproduce experimentally in flow networks. Natural flow networks like animal and plant vasculature can exhibit complex nonlinear behavior to facilitate fluid transport, so multistable flows may inform their functionality. To probe such phenomena in an analogous model system, we utilize an electronic network of hysteretic bistable resistors designed to have tunable negative differential resistivity. We demonstrate our system's capability to generate complex global memory states in the form of voltage patterns, which is mediated by the tunable nonlinearity of each element's current-voltage characteristic. We investigate avalanching behavior arising from effective interactions, and demonstrate how to encode explicit interactions of arbitrary form by taking advantage of the tunable circuitry design.

cond-mat.soft↗

arXiv2025

Dynamical loss functions shape landscape topography and improve learning in artificial neural networks

Dynamical loss functions are derived from standard loss functions used in supervised classification tasks, but are modified so that the contribution from each class periodically increases and decreases. These oscillations globally alter the loss landscape without affecting the global minima. In this paper, we demonstrate how to transform cross-entropy and mean squared error into dynamical loss functions. We begin by discussing the impact of increasing the size of the neural network or the learning rate on the depth and sharpness of the minima that the system explores. Building on this intuition, we propose several versions of dynamical loss functions and use a simple classification problem where we can show how they significantly improve validation accuracy for networks of varying sizes. Finally, we explore how the landscape of these dynamical loss functions evolves during training, highlighting the emergence of instabilities that may be linked to edge-of-instability minimization.

cs.LG↗

arXiv2025

Discovering dynamic laws from observations: the case of self-propelled, interacting colloids

Active matter spans a wide range of time and length scales, from groups of cells and synthetic self-propelled particles to schools of fish, flocks of birds, or even human crowds. The theoretical framework describing these systems has shown tremendous success at finding universal phenomenology. However, further progress is often burdened by the difficulty of determining the forces that control the dynamics of the individual elements within each system. Accessing this local information is key to understanding the physics dominating the system and to create the models that can explain the observed collective phenomena. In this work, we present a machine-learning model, a graph neural network, that uses the collective movement of the system to learn the active and two-body forces controlling the individual dynamics of the particles. We verify our approach using numerical simulations of active brownian particles, considering different interaction potentials and levels of activity. Finally, we apply our model to experiments of electrophoretic Janus particles, extracting the active and two-body forces that control the dynamics of the colloids. Due to this, we can uncover the physics dominating the behavior of the system. We extract an active force that depends on the electric field and also area fraction. We also discover a dependence of the two-body interaction with the electric field that leads us to propose that the dominant force between these colloids is a screened electrostatic interaction with a constant length scale. We expect that this methodology can open a new avenue for the study and modeling of experimental systems of active particles.

cond-mat.soft↗

arXiv2023

The fluidic memristor: collective phenomena in elastohydrodynamic networks

Fluid flow networks are ubiquitous and can be found in a broad range of contexts, from human-made systems such as water supply networks to living systems like animal and plant vasculature. In many cases, the elements forming these networks exhibit a highly non-linear pressure-flow relationship. Although we understand how these elements work individually, their collective behavior remains poorly understood. In this work, we combine experiments, theory, and numerical simulations to understand the main mechanisms underlying the collective behavior of soft flow networks with elements that exhibit negative differential resistance. Strikingly, our theoretical analysis and experiments reveal that a minimal network of nonlinear resistors, which we have termed a `fluidic memristor', displays history-dependent resistance. This new class of element can be understood as a collection of hysteresis loops that allows this fluidic system to store information. Our work provides insights that may inform new applications of fluid flow networks in soft materials science, biomedical settings, and soft robotics, and may also motivate new understanding of the flow networks involved in animal and plant physiology.

cond-mat.soft↗

arXiv2022

Model architecture can transform catastrophic forgetting into positive transfer

The work of McCloskey and Cohen popularized the concept of catastrophic interference. They used a neural network that tried to learn addition using two groups of examples as two different tasks. In their case, learning the second task rapidly deteriorated the acquired knowledge about the previous one. We hypothesize that this could be a symptom of a fundamental problem: addition is an algorithmic task that should not be learned through pattern recognition. Therefore, other model architectures better suited for this task would avoid catastrophic forgetting. We use a neural network with a different architecture that can be trained to recover the correct algorithm for the addition of binary numbers. This neural network includes conditional clauses that are naturally treated within the back-propagation algorithm. We test it in the setting proposed by McCloskey and Cohen and training on random additions one by one. The neural network not only does not suffer from catastrophic forgetting but it improves its predictive power on unseen pairs of numbers as training progresses. We also show that this is a robust effect, also present when averaging many simulations. This work emphasizes the importance that neural network architecture has for the emergence of catastrophic forgetting and introduces a neural network that is able to learn an algorithm.

cs.LG↗

arXiv2021

Tilting the playing field: Dynamical loss functions for machine learning

We show that learning can be improved by using loss functions that evolve cyclically during training to emphasize one class at a time. In underparameterized networks, such dynamical loss functions can lead to successful training for networks that fail to find a deep minima of the standard cross-entropy loss. In overparameterized networks, dynamical loss functions can lead to better generalization. Improvement arises from the interplay of the changing loss landscape with the dynamics of the system as it evolves to minimize the loss. In particular, as the loss function oscillates, instabilities develop in the form of bifurcation cascades, which we study using the Hessian and Neural Tangent Kernel. Valleys in the landscape widen and deepen, and then narrow and rise as the loss landscape changes during a cycle. As the landscape narrows, the learning rate becomes too large and the network becomes unstable and bounces around the valley. This process ultimately pushes the system into deeper and wider regions of the loss landscape and is characterized by decreasing eigenvalues of the Hessian. This results in better regularized models with improved generalization performance.

cs.LG↗

arXiv2021

Emergent dynamics in excitable flow systems

Flow networks can describe many natural and artificial systems. We present a model for a flow system that allows for volume accumulation, includes conduits with a non-linear relation between current and pressure difference, and can be applied to networks of arbitrary topology. The model displays complex dynamics, including self-sustained oscillations in the absence of any dynamics in the inputs and outputs. In this work we analytically show the origin of self-sustained oscillations for the 1D case. We numerically study the behavior of systems of arbitrary topology under different conditions: we discuss their excitability, the effect of different boundary conditions and wave propagation when the network has regions of conduits with linear conductance.

cond-mat.soft↗

arXiv2020

A numerical study of the rippling instability driven by electron-phonon coupling in graphene

Suspended graphene exhibits ripples of size ranging from 50 to 100 Å and height $\sim$10Å, however, their origin remains undetermined. Previous theoretical works have proposed that rippling in graphene might be generated by the coupling between the bending modes and the density of electrons. These studies theoretically proposed that in the thermodynamic limit a membrane of single layer graphene displays a lattice instability for large enough electron-phonon coupling, undergoing a phase transition from a flat phase to a rippled one. In this work we solve the elasticity equations for a finite membrane of graphene coupled to the charge distribution, modeled in the tight-binding approximation, we find that the electron-phonon coupling controls a transition from a stable flat configuration to a stable rippled phase.

cond-mat.mes-hall↗

arXiv2020

Buckling of thermalized elastic sheets

Steady progress in the miniaturization of structures and devices has reached a scale where thermal fluctuations become relevant and it is thus important to understand how such fluctuations affect their mechanical stability. Here, we investigate the buckling of thermalized sheets and we demonstrate that thermal fluctuations increase the critical buckling load due to the enhanced scale-dependent bending rigidity for sheets that are much larger than a characteristic thermal length scale. The presented results are universal and apply to a wide range of microscopic sheets. These results are especially relevant for atomically thin 2D materials, where thermal fluctuations can significantly increase the critical buckling load because the thermal length scale is on the order of nanometers at room temperature.

cond-mat.soft↗

arXiv2020

Topologically controlled emergent dynamics in flow networks

Flow networks are essential for both living organisms and enginneered systems. These networks often present complex dynamics controlled, at least in part, by their topology. Previous works have shown that topologically complex networks interconnecting explicitly oscillatory or excitable elements can display rich emerging dynamics. Here we present a model for complex flow networks with non-linear conductance that allows for internal accumulation/depletion of volume, without any inherent oscillatory or excitable behavior at the nodes. In the absence of any time dependence in the pressure input and output we observe emerging dynamics in the form of self-sustained waves, which travel through the system. The frequency of these waves depends strongly on the network architecture and it can be explained with a topological metric.

cond-mat.soft↗

arXiv2019

Tuning and jamming reduced to their minima

Inspired by protein folding, we introduce funnels into the complex cost function landscapes of two processes, the tuning of networks, and the jamming of ideal spheres. In both processes, geometrical frustration plays a role -- tuning pressure differences between pairs of target nodes far from the source in a flow network impedes tuning of nearby pairs more than the reverse process, while unjamming the system in one region can make it more difficult to unjam elsewhere. By modifying the cost functions to control the order in which functions are tuned or regions unjam, we smooth out local minima while leaving global minima unaffected, significantly increasing the success rate of reaching global minima.

cond-mat.soft↗

arXiv2019

Large-scale critical behavior of the rippling phase transition for graphene membranes

We analyze the spontaneous rippling of graphene membranes as function of the coupling between lattice deformations and electrons. We numerically study a model of an elastic membrane coupled to Dirac fermions. We identify a phase transition from a flat to a rippled configuration of the membrane when increasing the coupling and propose a scaling procedure that allows us to effectively reach arbitrary large system sizes. We find that the critical value of the coupling rapidly decays as the system increases its size, in agreement with the experimental observation of an unavoidable stable rippled state for suspended graphene membranes. This decay turns out to be controlled by a power law with a critical exponent $\sim 1/2$.

cond-mat.soft↗

arXiv2018

Thermal rectification with interacting electronic channels: Exploiting degeneracy, quantum superpositions and interference

This work explores different mechanisms that induce thermal rectification in the nanoscale. The presence of interacting energy channels combined with simple asymmetries is sufficient for promoting the desired behavior. We use simple quantum dot configurations, identifying the basic properties that enhance rectification for each case: the size of a quantum dot state space (which suggests the use of scaled up systems with many interacting channels), tunneling asymmetries due to coherent tunneling in a double quantum dot, or quantum interference in a triangular triple quantum dot. An efficient and tunable thermal diode is proposed using a channel capacitively coupled to a mesoscopic switch.

cond-mat.mes-hall↗

arXiv2017

Parameter dependence of high-frequency nonlinear oscillations and intrinsic chaos in short GaAs/(Al,Ga)As superlattices

We explore the design parameter space of short (5--25 period), n-doped, Ga/(Al,Ga)As semiconductor superlattices (SSLs) in the sequential resonant tunneling regime. We consider SSLs at cool (77K) and warm (295K) temperatures, simulating the electronic response to variations in (a) the number of SSL periods, (b) the contact conductivity, and (c) the strength of disorder (aperiodicities). Our analysis shows that the chaotic dynamical phases exist on a number of sub-manifolds of codimension zero within the design parameter space. This result provides an encouraging guide towards the experimental observation of high-frequency intrinsic dynamical chaos in shorter SSLs.

cond-mat.mes-hall↗

arXiv2017

Enhancing Chaotic Behavior at room temperature in GaAs/(Al,Ga)As Superlattices

Previous theoretical and experimental work has put forward 50-period semiconductor superlattices as fast, true random number generators at room temperature. Their randomness stems from feedback between nonlinear electronic dynamics and stochastic processes that are intrinsic to quantum transitions. This work theoretically demonstrates that shorter superlattices with higher potential barriers contain fully chaotic dynamics over several intervals of the applied bias voltage compared to the 50-periods device which presented a much weaker chaotic behavior. The chaos arises from deterministic dynamics, hence it persists even in the absence of additional stochastic processes. Moreover, the frequency of the chaotic current oscillations is higher for shorter superlattices. These features should allow for faster and more robust generation of true random numbers.

cond-mat.mes-hall↗

arXiv2017

Noise-enhanced chaos in a weakly coupled GaAs/(Al,Ga)As superlattice

Noise-enhanced chaos in a doped, weakly coupled GaAs/Al_{0.45}Ga_{0.55}As superlattice has been observed at room temperature in experiments as well as in the results of the simulation of nonlinear transport based on a discrete tunneling model. When external noise is added, both the measured and simulated current-versus-time traces contain irregularly spaced spikes for particular applied voltages, which separate a regime of periodic current oscillations from a region of no current oscillations at all. In the voltage region without current oscillations, the electric field profile consist of a low-field domain near the emitter contact separated by a domain wall consisting of a charge accumulation layer from a high-field regime closer to the collector contact. With increasing noise amplitude, spontaneous chaotic current oscillations appear over a wider bias voltage range. For these bias voltages, the domain boundary between the two electric-field domains becomes unstable, and very small current or voltage fluctuations can trigger the domain boundary to move toward the collector and induce chaotic current spikes. The experimentally observed features are qualitatively very well reproduced by the simulations. Increased noise can consequently enhance chaotic current oscillations in semiconductor superlattices.

nlin.CD↗

arXiv2016

STM driven transition from rippled to buckled graphene in a spin-membrane model

We consider a simple spin-membrane model for rippling in graphene. The model exhibits transitions from a flat but rippled membrane to a buckled one. At high temperature the transition is second order but it is first order at low temperature for appropriate strength of the spin-spin coupling. Driving the system across the first order phase transition in nonequilibrium conditions that mimic interaction of the graphene membrane with a STM tip explains recent experiments. In particular, we observe a reversible behavior for small values of the STM current and an irreversible transition from flat rippled membrane to rigid buckled membrane when the current surpasses a critical value. This work opens the possibility to test mechanical properties of graphene under different temperature and electrostatic conditions.

cond-mat.mes-hall↗