{"id":"8773c18e-d61f-4adf-8698-2700cdb9c61f","arxiv_id":"2505.03051","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Coupled nonchaotic Rulkov neurons on N-dimensional lattices exhibit dimension-dependent synchronization regimes, including synchronized hyperchaos and one-step lag synchronization at strong coupling.","lead":"This paper simulates lattices of simplified Rulkov neurons coupled in one to four spatial dimensions and maps how their spiking changes as electrical coupling strength grows. It identifies regimes such as synchronized hyperchaos, local quasi-bursting, and a checkerboard-style lag synchronization that appear when dimensions or connectivity increase.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Finite-time Lyapunov exponents (k=2000) plus 40-60 point moving-average smoothing leave the dimensional peak-shift and regime-boundary claims quantitatively unsupported.","rationale":"The reader's weakest assumption correctly identifies the finite-time Lyapunov exponent computation as the quantitative backbone. My reading of the manuscript reinforces this: the paper explicitly states that k=2000 introduces slightly more error, that some observed structure is a moving-average artifact, and that the small oscillations used to infer phase transitions are comparable to the variance of the Lyapunov exponent. The central claim concerning dimension-dependent shifts and regime sequences cannot be stronger than the numerical evidence supporting it. The reader's CONDITIONAL verdict already reflects this concern, so no verdict change is needed; the concrete check above would settle whether the concern actually lands.","tokens_in":27124,"tokens_out":5214,"duration_ms":60671,"concrete_test":"Recompute the lambda_1(g) curves in Fig. 12 for N=1-4 with zeta=4 using orbit lengths k=10^4 and k=10^5, with the moving-average smoothing removed, and repeat the heterogeneous cases with at least 5 independent random parameter draws. If the first-peak g-position shifts by more than the 0.02 smoothing window, or the peak height changes by more than ~0.01 across orbit lengths or realizations, the dimensional peak-shift claim is not yet established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claims, especially the rightward and upward shift of the first chaotic peak with dimension in Fig. 12, rest on Lyapunov exponents computed with orbits of only k=2000 timesteps and then smoothed with moving windows of 40-60 points over 1000 g values. The paper itself flags that this reduced orbit length 'introduces slightly more error,' attributes one visual feature to a moving-average artifact, and concedes that the small bumps used to infer miniature 'phase transitions' are 'on the same order of magnitude as the variance in the Lyapunov exponent.' Regime boundaries are assigned by visual inspection of these smoothed curves, and the heterogeneous cases use a single random realization with no error bars. If finite-time Lyapunov exponents have not converged, the sign and magnitude of lambda_1 can misclassify transient contraction as chaos, and peak positions extracted from smoothed curves can shift by up to about half the smoothing window (approximately 0.02 in g). Since the dimensional trend is the paper's headline finding, this numerical backbone is the most load-bearing assumption; without convergence evidence or smoothing-robustness checks, the trend could be an artifact.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript studies N-dimensional cubic lattices of electrically coupled nonchaotic Rulkov neurons (N = 1,...,4, plus large 2D and 3D cases) under nearest-neighbor and next-nearest-neighbor coupling, with homogeneous and heterogeneous neuron parameters. For each setting, the authors compute Lyapunov spectra via QR factorization using a piecewise Jacobian derived in Appendix A and classify dynamical regimes as a function of coupling strength g ∈ [0,1]. The main claims are: a universal sequence of regimes—uncoupled nonchaotic spiking, unsynchronized chaotic spiking, synchronized chaotic bursting, synchronized hyperchaos—in homogeneous lattices; an additional \"local quasi-bursting\" regime in low-conductance heterogeneous lattices; a delay and increase of the first chaos peak under NNN coupling due to \"destructive interference\"; miniature \"phase transitions\" in NNN and higher-dimensional lattices; a rightward and upward shift of the first chaotic peak with increasing spatial dimension (Fig. 12); and emergent local, quasi-, and lag synchronization in large lattices, including a one-timestep checkerboard lag-synchronized state at g = 1. The paper includes a tensorial formulation, a Jacobian derivation, and an implementation appendix.","tokens_in":27481,"tokens_out":5919,"duration_ms":57771,"significance":"If confirmed, the results would extend the study of Rulkov map networks from rings to higher-dimensional lattices and would provide a systematic regime catalog for a tractable discrete-time neuronal model. The manuscript has clear strengths: the Jacobian derivation in Appendix A is explicit and correctly reduces to the ring case; the code is publicly available; and the paper makes falsifiable predictions (e.g., peak-shift trends, disappearance of local quasi-bursting at higher dimension). However, the quantitative backbone—finite-time Lyapunov exponents with k = 2000, moving-window smoothing, visual regime assignment, and single realizations for heterogeneous cases—is not yet at the standard needed to support the headline dimensional trends. The contribution is potentially significant but requires substantial numerical validation.","major_comments":[{"comment":"The headline dimensional trend—the rightward and upward shift of the first chaotic peak in λ1 with increasing N—rests on Lyapunov exponents computed from orbits of only k = 2000 timesteps (Sec. 3.2, Fig. 12 caption) and on curves smoothed with moving windows of 40–60 points over 1000 values of g. The paper asserts that k = 2000 is long enough \"by computer experiment,\" but no convergence curves or error bars are shown. Because finite-time Lyapunov exponents can misclassify transient contraction as chaos and the moving average can shift peak locations by up to about half the window width (≈0.02–0.03 in g), the peak-shift trend and the associated predictions about higher dimensions are not yet quantitatively supported. Please provide convergence curves for representative g values and smoothing-robustness checks.","section":"§4.1, Fig. 12"},{"comment":"The miniature \"phase transitions\" are inferred from small bumps in the λ1 curves that the paper itself states are \"on the same order of magnitude as the variance in the Lyapunov exponent\" (Sec. 3.2). The supporting visualization (Fig. 10) shows six neurons at six g values, but it does not establish that the bumps are statistically distinct from fluctuations or that they arise from coordinated threshold crossings. Without estimates of the Lyapunov-exponent variance or repeated realizations, this claim is not supported.","section":"§3.2, Figs. 8 and 10"},{"comment":"For the heterogeneous cases, a single random realization is used for each parameter distribution (Sec. 2.2). Consequently, statements such as \"the local quasi-bursting phase in the N = 2 lattice has been extinguished\" in the fully heterogeneous case (Sec. 4.1) or the comparison between partially and fully heterogeneous curves in Figs. 4b and 4c may be realization-specific. Ensemble averages over multiple draws of σi and αi, with error bars, are needed before these contrasts can be taken as properties of the distribution.","section":"§2.2, Figs. 4, 8, 12"},{"comment":"The regime classification itself is performed by visual inspection of six adjacent neurons and representative snapshots (e.g., Figs. 5–7, 10, 13). Terms such as \"local quasi-bursting\" and boundaries like 0.2 ≲ g ≲ 0.8 are not tied to quantitative order parameters (e.g., synchronization error, burst length, fraction of silent neurons). Because the paper's central claims concern the existence and ordering of these regimes, I would ask for quantitative definitions and validation, especially for the newly introduced \"local quasi-bursting\" regime.","section":"§4.1, Fig. 12a"}],"minor_comments":[{"comment":"The caption says \"three select values of g\" but the figure contains six panels; please correct the caption.","section":"Fig. 10 caption"},{"comment":"The notation t = k for discrete time is later reused for the Lyapunov orbit length k; this is not an error but could be clarified to avoid confusion in Sec. 3.2.","section":"Sec. 2.1"},{"comment":"Reference [45] appears to misstate the title of the McCulloch–Pitts paper; the canonical title is \"A logical calculus of the ideas immanent in nervous activity.\"","section":"Ref. [45]"},{"comment":"The claim that λ1 \"increases monotonically with dimension\" around g ≈ 0.2 and g ≳ 0.9 is not obvious from the overlaid curves in Fig. 12a because the N = 2, 3, 4 curves nearly overlap; a zoomed inset or separate panels would help substantiate this statement.","section":"Sec. 4.1"}],"recommendation":"major_revision","confidential_remarks":"The paper's main numerical risk is the reliance on finite-time Lyapunov exponents without convergence evidence; the requested convergence curves and ensemble statistics are essential before publication. The authors should also clarify the status of Ref. [36], which appears to be an unpublished preprint by the same lead author and is used as the validation baseline for the N = 1 ring case."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this is an honest and mostly competent computational extension of the 1D ring-lattice Rulkov work to N-dimensional cubic lattices. The main new claim — a regular rightward and upward shift of the first chaotic peak in λ1 as N increases — is plausible, but the numerical evidence as presented does not yet make it solid.\n\nWhat the paper does well: it generalizes Ref. [36] cleanly, gives an explicit Jacobian derivation in Appendix A that reduces to the ring case when N=1, and the code and data are on GitHub. The authors also flag their own limitations: they state that k=2000 orbits 'introduce slightly more error', note that the 'miniature phase transitions' bumps are 'on the same order of magnitude as the variance' in λ1, and even attribute one visual feature to the moving-average smoothing. That candor is real.\n\nThe soft spots are the quantitative backbone. Lyapunov exponents are computed from finite orbits of 2000 steps (10000 for the 2D NN case), then smoothed with moving windows of 40–60 points over 1000 g-values. There are no convergence curves, no error bars, and regime boundaries are assigned by eye. The heterogeneous cases use a single random realization. The moving-average smoothing alone could shift peak locations by roughly half the window, about 0.02 in g. Since the dimensional trend in Fig. 12 is the paper's headline finding, this is load-bearing, not a minor detail. The more speculative mechanisms — 'destructive interference' and 'miniature phase transitions' — are asserted rather than measured, and the paper's own caveat about the bump size weakens the latter.\n\nThat said, I don't think the work is unsound. The qualitative regime sequence (nonchaotic spiking → unsynchronized chaotic spiking → synchronized chaotic bursting → synchronized hyperchaos) is consistent across dimensions and matches the known 1D behavior. The Jacobian is right, and the large-lattice observations (radial spreading, checkerboard lag synchronization) are visually documented and plausible. The issues are about quantitative support, not about the model being wrong.\n\nWho should read this: anyone working on map-based neuron networks or lattice synchronization. It is a useful extension with a clear derivation and an honest limitations section, but the central trend claims need stronger numerical validation before they should be cited as established.\n\nRecommendation: send it to peer review, but with a request for major revisions. The authors should add convergence evidence for the Lyapunov exponents, error bars or multiple realizations, and either quantitative regime-boundary detection or appropriately softened claims. The core material is worth referee time.","headline":"Honest N-dimensional Rulkov lattice extension; the dimensional peak-shift trend is plausible but rests on finite-time Lyapunov exponents without convergence evidence or error bars.","tokens_in":27914,"tokens_out":2712,"would_cite":true,"duration_ms":25940,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37D45","37M25","37N25"],"pacs":["05.45.-a","05.45.Xt"],"model":"deepseek-v4-flash","headline":"As electrical coupling grows, nonchaotic Rulkov neurons on N-dimensional lattices follow a single regime sequence from independent spiking to synchronized hyperchaos, with dimensionality shifting where chaos first appears.","keywords":["Rulkov neurons","N-dimensional lattices","hyperchaos","Lyapunov exponents","synchronized bursting","lag synchronization","electrical coupling","next-nearest-neighbor coupling"],"falsifier":"Recompute the maximal Lyapunov exponent sweep for the $N=2$ homogeneous nearest-neighbor lattice with $\\zeta=8$ using orbit length $k=100000$ and fifty independent initial conditions; if the sharp drop near $g\\approx 0.2$, the rightward peak shift with $N$, and the small bumps in the next-nearest-neighbor curves do not survive these longer runs, the central regime-transition claims are called into question.","tokens_in":1916,"feed_emoji":"🧠","tokens_out":3385,"duration_ms":114122,"temperature":0.7,"pith_summary":"The paper sets out to show that a minimal discrete-time neuron model, the nonchaotic Rulkov map, placed on cubic lattices in one to four spatial dimensions, develops a reproducible sequence of collective dynamical regimes as electrical coupling strength $g$ is swept from 0 to 1. In homogeneous lattices the sequence runs from uncoupled nonchaotic spiking through unsynchronized chaotic spiking and synchronized chaotic bursting to synchronized hyperchaos; heterogeneous lattices add a low-coupling 'local quasi-bursting' phase. Dimensionality and connectivity do not alter the sequence itself, but the paper argues they shift its thresholds: the first chaotic peak in the maximal Lyapunov exponent moves to larger $g$ and higher values as $N$ grows, and next-nearest-neighbor coupling delays the onset of synchronized bursting through a diluting 'destructive interference' of many neighbors. At maximal coupling, large 2D and 3D lattices display a checkerboard of adjacent neurons alternating high and low voltage each timestep, an extreme form of one-timestep lag synchronization. A reader should care because this suggests that spatial wiring, not the internal complexity of the neuron model, can generate dimension-dependent chaos and synchronization, and it gives a computationally cheap test bed for synchronization phenomena.","feed_headline":"Rulkov neuron lattices hit synchronized hyperchaos in every tested dimension","feed_subtitle":"Sweeping electrical coupling from 0 to 1 pushes 1D-4D lattices through the same chaos sequence, ending in a checkerboard lag-sync state.","key_machinery":"The carrying object is the lattice state tensor $X_{ia}=(x_i,y_i)$ of Rulkov maps with electrical coupling $C_i(k)=\\frac{g}{|N_i|}\\sum_{j\\in N_i}(x_j-x_i)$, where $|N_i|=2N$ for nearest neighbors and $4N$ for next-nearest neighbors, with periodic boundary conditions. The iteration function feeds this current into both the fast variable (through an effective shift of the slow variable) and the slow variable (through an effective shift of the excitation parameter), which is the mechanism behind the high-frequency spiking and subsequent quiescence that defines synchronized chaotic bursting. The argument is carried quantitatively by an explicitly derived tensorial Jacobian, converted to a matrix via a base-$\\zeta$ index map, whose QR-based Lyapunov spectra provide the regime boundaries. Two mechanisms explain the key patterns: 'destructive interference' dilutes each pairwise interaction when a neuron has more neighbors, delaying synchronization, and the discrete-time lag in current flow reverses the voltage difference every timestep at $g=1$, producing checkerboard lag synchronization.","core_discovery":"The central claim is that electrically coupled nonchaotic Rulkov neurons on an $N$-dimensional cubic lattice, with nearest-neighbor or next-nearest-neighbor coupling, pass through the same basic sequence of regimes as the coupling conductance $g$ is increased. For homogeneous parameters the regimes are uncoupled nonchaotic spiking, unsynchronized chaotic spiking, synchronized chaotic bursting, and synchronized hyperchaos; for heterogeneous parameters, a local quasi-bursting phase appears at low coupling. The paper's quantitative support is the maximal Lyapunov exponent $\\lambda_1$ computed from the full Jacobian: as $N$ increases from 1 to 4, the first chaotic peak shifts rightward and upward, the descent into synchronized bursting becomes more uniform, and the final rise to synchronized hyperchaos near $g=1$ remains sharp. In next-nearest-neighbor lattices the synchronized-hyperchaos rise disappears within $g\\in[0,1]$, which the paper attributes to destructive interference from the larger neighborhood. At $g=1$ in large 2D and 3D lattices, the paper identifies checkerboard antiphase spiking of adjacent neurons as extreme one-timestep lag synchronization.","pith_inferences":["If the rightward peak shift is a genuine scaling law, then the transition coupling $g_c(N)$ should either approach a finite limit or follow a dependence like $g_c\\sim 1/(2N)$ for large $N$; computing $N=5$ and $N=6$ with longer orbits would test this prediction.","The checkerboard one-timestep lag synchronization is a discrete-time analogue of anti-phase synchronization; in continuous-time models it may appear as half-period lag, giving a concrete signature to look for in coupled oscillator experiments or neuromorphic circuits.","The local quasi-bursting phase in heterogeneous lattices may be a finite-size or finite-time artifact: its valley moves toward $g=0$ as $N$ grows, so larger systems could show a monotonic chaos onset instead.","The paper's 'universal saturation' in strongly coupled heterogeneous lattices suggests the hyperchaotic dynamics may become extensive with dimension; computing the Kaplan-Yorke dimension per neuron across $N$ could reveal a dimension-independent density of positive Lyapunov directions."],"forward_implications":["In NN-coupled homogeneous lattices, the same four regimes appear for $N=1$ through $N=4$, so the regime sequence is a robust feature of the lattice rather than an accident of one dimension.","Because the first chaotic peak shifts rightward and upward with $N$, higher-dimensional lattices remain unsynchronized-chaotic over a wider coupling window and reach stronger chaos there; extrapolating the trend predicts sharper transitions in higher dimensions.","Next-nearest-neighbor coupling eliminates the synchronized-hyperchaos regime for $g\\in[0,1]$; the lattice stays in synchronized chaotic bursting even at $g=1$.","Large 2D and 3D lattices show quasi-synchronization and localized synchronization, and at $g=1$ they display extreme one-timestep lag synchronization as checkerboard anti-phase spiking.","Heterogeneity can create local quasi-bursting at low coupling, but the phase weakens as dimension increases and appears extinguished for fully heterogeneous lattices in $N=1$ and $N=2$."],"supporting_citations":[{"why":"Defines the nonchaotic Rulkov map and its spiking-bursting regimes, the single-neuron substrate of the lattice.","marker":"[15]"},{"why":"Supplies the one-dimensional ring-lattice baseline and comparison for the regime sequence and hyperchaotic analysis.","marker":"[36]"},{"why":"Accounts for how bursting and chaotic dynamics arise in individual Rulkov neurons and is used for Lyapunov analysis.","marker":"[39]"},{"why":"Provides the QR-factorization method used to compute the Lyapunov spectra.","marker":"[44]"},{"why":"Defines lag synchronization in time-delayed systems, the basis for the checkerboard interpretation.","marker":"[55]"},{"why":"Introduces lag synchronization in coupled chaotic oscillators, used to identify the one-timestep lag regime.","marker":"[56]"},{"why":"Defines quasi-synchronization with error bound, used to characterize large-lattice collective behavior.","marker":"[50]"},{"why":"Defines lag quasi-synchronization under parameter mismatch, supporting the quasi-synchronized description.","marker":"[51]"}],"fun_headline_variants":["Hyperchaos syncs Rulkov lattices across 1D to 4D","N-D Rulkov lattices end in synchronized hyperchaos","Same chaos sequence: Rulkov lattices in 1D to 4D hyperchaos"],"cache_read_input_tokens":30080,"weakest_assumption_plain":"The regime classification and all transition thresholds rest on Lyapunov exponents computed from short finite orbits ($k=2000$, or $10000$ for the two-dimensional nearest-neighbor case) with no convergence curves or error bars, so numerical error could shift the reported regime boundaries.","fun_headline_variants_meta":{"raw":{"variants":["Hyperchaos syncs Rulkov lattices across 1D to 4D","N-D Rulkov lattices end in synchronized hyperchaos","Same chaos sequence: Rulkov lattices in 1D to 4D hyperchaos"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000947,"raw_usage":{"total_tokens":4062,"prompt_tokens":983,"completion_tokens":3079,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":599,"completion_tokens_details":{"reasoning_tokens":3008}},"tokens_in":599,"tokens_out":3079,"duration_ms":22489,"temperature":1.0,"reasoning_tokens":3008,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:00:51.513740+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the maximal Lyapunov exponent sweep for the $N=2$ homogeneous nearest-neighbor lattice with $\\zeta=8$ using orbit length $k=100000$ and fifty independent initial conditions; if the sharp drop near $g\\approx 0.2$, the rightward peak shift with $N$, and the small bumps in the next-nearest-neighbor curves do not survive these longer runs, the central regime-transition claims are called into question.","supporting_citations":[{"cited_title":"Modeling of spiking-bursting neural behavior using two-dimensional map,","cited_arxiv_id":null,"evidence_quote":"Defines the nonchaotic Rulkov map and its spiking-bursting regimes, the single-neuron substrate of the lattice."},{"cited_title":"Ergodic theory of chaos and strange attractors,","cited_arxiv_id":null,"evidence_quote":"Provides the QR-factorization method used to compute the Lyapunov spectra."},{"cited_title":"Lag synchronization in time-delayed systems,","cited_arxiv_id":null,"evidence_quote":"Defines lag synchronization in time-delayed systems, the basis for the checkerboard interpretation."},{"cited_title":"From phase to lag synchronization in coupled chaotic oscillators,","cited_arxiv_id":null,"evidence_quote":"Introduces lag synchronization in coupled chaotic oscillators, used to identify the one-timestep lag regime."},{"cited_title":"Quasi-synchronization of heterogeneous dynamic networks via distributed impulsive control: Error estimation, optimization and design,","cited_arxiv_id":null,"evidence_quote":"Defines quasi-synchronization with error bound, used to characterize large-lattice collective behavior."},{"cited_title":"Lag quasi-synchronization of coupled delayed systems with parameter mismatch,","cited_arxiv_id":null,"evidence_quote":"Defines lag quasi-synchronization under parameter mismatch, supporting the quasi-synchronized description."}],"review_version":1}