Disordered Systems and Neural Networks
See recent articles
Showing new listings for Wednesday, 12 March 2025
- [1] arXiv:2503.07705 [pdf, html, other]
-
Title: Geometric Delocalization in Two DimensionsComments: 4.5 pages + 5 figures + supplemental materialSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Mathematical Physics (math-ph); Quantum Physics (quant-ph)
We demonstrate the existence of transient two-dimensional surfaces where a random-walking particle escapes to infinity in contrast to localization in standard flat 2D space. We first prove that any rotationally symmetric 2D membrane embedded in flat 3D space cannot be transient. Then we formulate a criterion for the transience of a general asymmetric 2D membrane. We use it to explicitly construct a class of transient 2D manifolds with a non-trivial metric and height function but ``zero average curvature,'' which we dub tablecloth manifolds. The absence of the logarithmic infrared divergence of the Laplace-Beltrami operator in turn implies the absence of weak localization, non-existence of bound states in shallow potentials, and breakdown of the Mermin-Wagner theorem and Kosterlitz-Thouless transition on the tablecloth manifolds, which may be realizable in both quantum simulators and corrugated two-dimensional materials.
- [2] arXiv:2503.07786 [pdf, html, other]
-
Title: Effect of imaginary gauge on wave transport in driven-dissipative systemsComments: 13 pages, 9 figuresSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Optics (physics.optics)
Wave transport in disordered media is a fundamental problem with direct implications in condensed matter, materials science, optics, atomic physics, and even biology. The majority of studies are focused on Hermitian systems to understand disorder-induced localization. However, recent studies of non-Hermitian disordered media have revealed unique behaviors, with a universal principle emerging that links the eigenvalue spectrum of the disordered Hamiltonian and its statistics with its transport properties. In this work we show that the situation can be very different in driven-dissipative lattices of cavities, where a uniform gain applied equally to all the components of the system can act as a knob for controlling the wave transport properties without altering the eigenvalue statistics of the underlying Hamiltonian. Our results open a new avenue for developing a deeper insight into the transport properties in disordered media and will aid in building new devices as well. Our work which is presented in the context of optics generalizes to any physical platforms where gain can be implemented. These include acoustics, electronics, and coupled quantum oscillators such as atoms, diamond centers and superconducting qubits.
- [3] arXiv:2503.07796 [pdf, html, other]
-
Title: Topological mechanical neural networks as classifiers through in situ backpropagation learningComments: 26 pages, 5 figuresSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn)
Recently, a new frontier in computing has emerged with physical neural networks(PNNs) harnessing intrinsic physical processes for learning. Here, we explore topological mechanical neural networks(TMNNs) inspired by the quantum spin Hall effect(QSHE) in topological metamaterials, for machine learning classification tasks. TMNNs utilize pseudospin states and the robustness of the QSHE, making them damage-tolerant for binary classification. We first demonstrate data clustering using untrained TMNNs. Then, for specific tasks, we derive an in situ backpropagation algorithm - a two-step, local-rule method that updates TMNNs using only local information, enabling in situ physical learning. TMNNs achieve high accuracy in classifications of Iris flowers, Penguins, and Seeds while maintaining robustness against bond pruning. Furthermore, we demonstrate parallel classification via frequency-division multiplexing, assigning different tasks to distinct frequencies for enhanced efficiency. Our work introduces in situ backpropagation for wave-based mechanical neural networks and positions TMNNs as promising neuromorphic computing hardware for classification tasks.
- [4] arXiv:2503.08143 [pdf, html, other]
-
Title: Phase Transitions and Critical Behavior in Quasi-One-Dimensional Two-Channel Systems with Quasiperiodic DisorderJournal-ref: Phys. Scr. 100 035917 (2025)Subjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Strongly Correlated Electrons (cond-mat.str-el)
We investigate the localization properties of a quasi-one-dimensional two-channel system with symmetric and asymmetric onsite energies using the Aubry-André model. By analyzing the Lyapunov exponent and localization length, we characterize the phase transitions and critical behavior of the system. For the symmetric model, we obtain the phase diagram for the entire spectrum, revealing mobility edges between delocalized and localized states. In contrast, for the asymmetric model, we identify a critical line $ \lambda_1^c + \lambda_2^c \approx 0.5 $ marking the phase transition between delocalized and localized states. We also study the effects of the inter-channel coupling $ \tilde{t} $, and observe that increasing $ \tilde{t} $ reduces the delocalized phase space, shifting the transition from $ \lambda_1 = \lambda_2 = 1 $ at $ \tilde{t} = 0 $ to $ \lambda_1^c + \lambda_2^c \approx 0.5 $ at larger $ \tilde{t} $. Furthermore, the phase transition point is found to be sensitive to both $ \tilde{t} $ and the incommensurate modulation parameter $ b $. While the general phase transition behavior is preserved, subtle differences arise for different values of $ b $, indicating a dependence of the phase boundary on both parameters. Using the cost function approach, we calculate the critical potential strength $ \lambda_c $ and the critical exponent $ \nu $, with $ \nu \approx 0.5 $ for the middle of the spectrum in both symmetric and asymmetric models.
- [5] arXiv:2503.08194 [pdf, html, other]
-
Title: Directional Localization in Disordered 2D Tight-Binding Systems: Insights from Single Particle Entanglement MeasuresJournal-ref: Eur. Phys. J. B, 97, 18 (2024)Subjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Strongly Correlated Electrons (cond-mat.str-el)
We investigate the directional localization properties of wave-functions in a two-dimensional tight-binding model with uniform hopping and correlated random on-site energies. By controlling the disorder correlation strength with a parameter $\alpha$, we explore the effects of disorder on wave-function localization using Single Particle Entanglement Entropy (SPEE) and Single Particle Rényi Entropy (SPRE) at different values of $q$. Our analysis includes two distinct randomness structures: row-wise and fully correlated disorder. We find that row-wise disorder maintains maximal entanglement for horizontal cuts while enhancing horizontal spread for vertical cuts as $\alpha$ increases. In contrast, fully correlated disorder leads to reduced vertical entanglement for horizontal cuts and increased horizontal entanglement for vertical cuts with rising $\alpha$. Additionally, our results show that the difference between SPEE and SPRE provides valuable insights into localization behavior. These findings highlight the significance of directional properties in understanding localization transitions in disordered systems.
- [6] arXiv:2503.08266 [pdf, html, other]
-
Title: Sampling the space of solutions of an artificial neural networkComments: 27 pages, 11 figuresSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Statistical Mechanics (cond-mat.stat-mech); Probability (math.PR)
The weight space of an artificial neural network can be systematically explored using tools from statistical mechanics. We employ a combination of a hybrid Monte Carlo algorithm which performs long exploration steps, a ratchet-based algorithm to investigate connectivity paths, and coupled replica models simulations to study subdominant flat regions. Our analysis focuses on one hidden layer networks and spans a range of energy levels and constrained density regimes. Near the interpolation threshold, the low-energy manifold shows a spiky topology. While these spikes aid gradient descent, they can trap general sampling algorithms at low temperatures; they remain connected by complex paths in a confined, non-flat region. In the overparameterized regime, however, the low-energy manifold becomes entirely flat, forming an extended complex structure that is easy to sample. These numerical results are supported by an analytical study of the training error landscape, and we show numerically that the qualitative features of the loss landscape are robust across different data structures. Our study aims to provide new methodological insights for developing scalable methods for large networks.
New submissions (showing 6 of 6 entries)
- [7] arXiv:2503.08069 (cross-list from q-bio.NC) [pdf, html, other]
-
Title: Freezing chaos without synaptic plasticityComments: 19 pages, 9 figuresSubjects: Neurons and Cognition (q-bio.NC); Disordered Systems and Neural Networks (cond-mat.dis-nn); Neural and Evolutionary Computing (cs.NE); Chaotic Dynamics (nlin.CD)
Chaos is ubiquitous in high-dimensional neural dynamics. A strong chaotic fluctuation may be harmful to information processing. A traditional way to mitigate this issue is to introduce Hebbian plasticity, which can stabilize the dynamics. Here, we introduce another distinct way without synaptic plasticity. An Onsager reaction term due to the feedback of the neuron itself is added to the vanilla recurrent dynamics, making the driving force a gradient form. The original unstable fixed points supporting the chaotic fluctuation can then be approached by further decreasing the kinetic energy of the dynamics. We show that this freezing effect also holds in more biologically realistic networks, such as those composed of excitatory and inhibitory neurons. The gradient dynamics are also useful for computational tasks such as recalling or predicting external time-dependent stimuli.
Cross submissions (showing 1 of 1 entries)
- [8] arXiv:2406.12677 (replaced) [pdf, html, other]
-
Title: Scaling of the Integrated Quantum Metric in Disordered Topological PhasesComments: 6 pages, 3 figuresSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Mesoscale and Nanoscale Physics (cond-mat.mes-hall)
We report a study of a disorder-dependent real-space representation of the quantum geometry in topological systems. Thanks to the development of an efficient linear-scaling numerical methodology based on the kernel polynomial method, we can explore nontrivial behavior of the integrated quantum metric and Chern number in disordered systems with sizes reaching the experimental scale. We illustrate this approach in the disordered Haldane model, examining the impact of Anderson disorder and vacancies on the trivial and topological phases captured by this model.
- [9] arXiv:2409.12781 (replaced) [pdf, html, other]
-
Title: On the topology of solutions to random continuous constraint satisfaction problemsSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Statistical Mechanics (cond-mat.stat-mech)
We consider the set of solutions to $M$ random polynomial equations whose $N$ variables are restricted to the $(N-1)$-sphere. Each equation has independent Gaussian coefficients and a target value $V_0$. When solutions exist, they form a manifold. We compute the average Euler characteristic of this manifold in the limit of large $N$, and find different behavior depending on the target value $V_0$, the ratio $\alpha=M/N$, and the variances of the coefficients. We divide this behavior into five phases with different implications for the topology of the solution manifold. When $M=1$ there is a correspondence between this problem and level sets of the energy in the spherical spin glasses. We conjecture that the transition energy dividing two of the topological phases corresponds to the energy asymptotically reached by gradient descent from a random initial condition, possibly resolving an open problem in out-of-equilibrium dynamics. However, the quality of the available data leaves the question open for now.
- [10] arXiv:2410.18010 (replaced) [pdf, other]
-
Title: Triplets of local minima in a high-dimensional random landscape: Correlations, clustering, and memoryless activated jumpsSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Statistical Mechanics (cond-mat.stat-mech)
We compute the distribution of triplets of stationary points in the energy landscape of the spherical p-spin model, by evaluating the quenched three-point complexity by means of the Kac-Rice formalism. We show the occurrence of transitions in the organization of stationary points in the landscape, identifying regions where local minima and saddles accumulate and cluster around other stationary points, thus displaying the presence of correlations in the landscape. We discuss the implications of these findings for the dynamical exploration of the energy landscape in the activated regime, specifying conditions under which transitions between local minima are expected to exhibit correlated rates and when, conversely, activated jumps are likely to be memoryless.
- [11] arXiv:2502.08192 (replaced) [pdf, html, other]
-
Title: Probing the many-body localized spin-glass phase through quench dynamicsSubjects: Disordered Systems and Neural Networks (cond-mat.dis-nn); Quantum Gases (cond-mat.quant-gas); Strongly Correlated Electrons (cond-mat.str-el); Quantum Physics (quant-ph)
Eigenstates of quantum many-body systems are often used to define phases of matter in and out of equilibrium; however, experimentally accessing highly excited eigenstates is a challenging task, calling for alternative strategies to dynamically probe nonequilibrium phases. In this work, we characterize the dynamical properties of a disordered spin chain, focusing on the spin-glass regime. Using tensor-network simulations, we observe oscillatory behavior of local expectation values and bipartite entanglement entropy. We explain these oscillations deep in the many-body localized spin glass regime via a simple theoretical model. From perturbation theory, we predict the timescales up to which our analytical description is valid and confirm it with numerical simulations. Finally, we study the correlation length dynamics, which, after a long-time plateau, resumes growing in line with renormalization group (RG) expectations. Our work suggests that RG predictions can be quantitatively tested against numerical simulations and experiments, potentially enabling microscopic descriptions of dynamical phases in large systems.
- [12] arXiv:2407.16544 (replaced) [pdf, html, other]
-
Title: Real-Time Out-of-Equilibrium Quantum Dynamics in Disordered MaterialsSubjects: Mesoscale and Nanoscale Physics (cond-mat.mes-hall); Disordered Systems and Neural Networks (cond-mat.dis-nn); Materials Science (cond-mat.mtrl-sci)
We report a linear-scaling numerical method for exploring nonequilibrium electron dynamics in systems of arbitrary complexity. Based on the Chebyshev expansion of the time evolution of the single-particle density matrix, the method gives access to nonperturbative excitation and relaxation phenomena in models of disordered materials with sizes on the experimental scale. After validating the method by applying it to saturable optical absorption in clean graphene, we uncover that disorder can enhance absorption in graphene and that the interplay between light, anisotropy, and disorder in nanoporous graphene might be appealing for sensing applications. Beyond the optical properties of graphene-like materials, the method can be applied to a wide range of large-area materials and systems with arbitrary descriptions of defects and disorder.
- [13] arXiv:2407.20631 (replaced) [pdf, html, other]
-
Title: Hedgehog topological defects in 3D amorphous solidsComments: v2: revised version, geometrical characterization of the topological defects addedSubjects: Soft Condensed Matter (cond-mat.soft); Disordered Systems and Neural Networks (cond-mat.dis-nn); Materials Science (cond-mat.mtrl-sci); Statistical Mechanics (cond-mat.stat-mech)
The underlying structural disorder renders the concept of topological defects in amorphous solids difficult to apply and hinders a first-principle identification of the microscopic carriers of plasticity and of the regions more prone to structural rearrangements (``soft spots''). Recently, it has been proposed that well-defined topological defects can still be identified in glasses, and correlated to local and global plasticity, by looking at the eigenvector field or the particle displacement field. Nevertheless, all the existing proposals and analyses are only valid in two spatial dimensions. In this work, we propose the idea of using hedgehog topological defects to characterize the plasticity of 3D glasses and to geometrically predict the location of their soft spots. We corroborate our proposal by simulating a Kremer-Grest 3D polymer glass, and by using both the normal mode eigenvector field and the displacement field around large plastic events. Contrary to the 2D case, the sign of the topological charge defined from the eigenvector field is ambiguous and the geometry of the topological defects, whether radial or hyperbolic, plays a fundamental role in 3D. In fact, we find that the topological hedgehog defects relevant for plasticity are those exhibiting hyperbolic geometry, resembling the saddle-point structure of 2D topological defects with negative winding number (anti-vortices). Our results confirm that a topological characterization of plasticity in glasses is feasible and provide a concrete realization of this program in 3D amorphous systems.
- [14] arXiv:2409.10437 (replaced) [pdf, html, other]
-
Title: Color symmetry breaking in the Potts spin glassComments: 6 pages, final versionSubjects: Probability (math.PR); Disordered Systems and Neural Networks (cond-mat.dis-nn)
The Potts spin glass is an analogue of the Sherrington-Kirkpatrick model in which each spin can take one of $\kappa$ possible values, which we interpret as colors. It was suggested in arXiv:2310.06745 that the order parameter for this model is always invariant with respect to permutations of the colors. We show here that this is false whenever $\kappa \ge 58$.
- [15] arXiv:2503.07031 (replaced) [pdf, other]
-
Title: Negative Local Partial Density of StatesSubjects: Quantum Physics (quant-ph); Disordered Systems and Neural Networks (cond-mat.dis-nn)
Real quantum systems can exhibit a local object called local partial density of states (LPDOS) that cannot be proved within the axiomatic approach of quantum mechanics. We demonstrate that real mesoscopic system that can exhibit Fano resonances will show this object and also very counterintuitively it can become negative, resulting in the enhancement of coherent currents.