arXiv subjects

Nalin Dhiman

Publications and source records attributed to Nalin Dhiman.

5 recordsLinked to original sources

arXiv2026

Topological Sensitivity in Connectome-Constrained Neural Networks

Connectome-constrained neural networks are often evaluated against sparse random controls and then interpreted as evidence that biological graph topology improves learning efficiency. We revisit that claim in a controlled flyvis-based study using a Drosophila connectome, a naive self-loop-matched random graph, and a degree-preserving rewired null. Under weak controls, in which both models were recovered from a connectome-trained checkpoint and the null matched only global graph counts, the connectome appeared substantially better in early loss, mean activity, and runtime. That picture changed under stricter controls. Training both graphs from a shared random initialization removed the early loss advantage, and replacing the naive null by a degree-preserving null removed the apparent activity advantage. A five-sample degree-preserving ensemble and a pre-training activity-scale diagnostic further strengthened this revised interpretation. We also report a descriptive mechanism analysis of the earlier weak-control comparison, but we treat it as behavioral characterization rather than proof of causal superiority. We show that previously reported topology advantages in connectome-constrained neural networks can arise from initialization and null-model confounds, and largely disappear under fair from-scratch initialization and degree-preserving controls.

q-bio.NC↗

arXiv2026

The Physics of the Dancing \emph{Deity}: Coupled Oscillators in Himalayan Processions

In parts of Himachal Pradesh (Kullu and Mandi) and the Western Himalaya, village deities (\emph{devt\=a}) are carried through the landscape on shoulder-borne palanquins or ``raths.'' Participants often describe these raths as agents: they \emph{choose} routes, signal assent or refusal, and sometimes ``move on their own'' as if people are not moving them but are instead being moved. This paper offers : (i) a mechanistic model in which a palanquin interacts with human carriers modeled as coupled limit-cycle oscillators, and (ii) a philosophical analysis of how music and gurus/oracular specialists (\emph{g\=ur}/``guru'' in local English) function as couplings that stabilize collective interpretation, producing what we call \emph{distributed agency}. On the physics side we build a six-degree-of-freedom rigid-body model with unilateral handle contacts, base excitation from walking, and a Kuramoto-Adler phase description for interpersonal coupling and musical entrainment. We prove standard phase-locking conditions (Adler-type capture range) and show how unilateral contact can rectify periodic forcing and inject harmonics, creating parameter regimes in which near-perfect synchrony produces large-amplitude roll. On the simulation side we report an ensemble study (30 seeds per condition) from an archived ``Palanquin Simulator'' package: a ``baseline'' condition produces small roll ($\mathrm{RMS}\approx 0.15^{\circ}$) and moderate synchrony, whereas a ``music'' entrainment condition produces near-unity synchrony ($\approx 0.99$) but also robust roll instability ($\mathrm{RMS}\approx 18^{\circ}$) and frequent contact loss. We do \emph{not} claim to have validated the model against field data; we treat the simulations as a \emph{proof of plausibility} and as a generator of falsifiable predictions.

physics.soc-ph↗

arXiv2026

sidmkit: A Reproducible Toolkit for SIDM Phenomenology and Galaxy Rotation-Curve Modeling

Self-interacting dark matter (SIDM) is a well-motivated extension of cold dark matter that can modify halo structure on galactic and group scales while remaining consistent with large-scale structure. However, practical SIDM work often requires bridging several layers, including microphysical scattering models, velocity-dependent effective cross sections, phenomenological astrophysical constraints, and (separately) data-driven halo fits, such as rotation curves. In this paper, we describe \texttt{sidmkit}, a transparent and reproducible Python package designed to support SIDM ``micro$\rightarrow$macro'' calculations and to provide a robust batch pipeline for fitting rotation curves in the SPARC data. On the SIDM side, \texttt{sidmkit} implements velocity-dependent momentum-transfer cross sections for a Yukawa interaction using standard analytic approximations (Born, classical, and Hulth\'en-based) with a numerical partial-wave option for spot checks. It also provides consistent velocity-moment averaging for Maxwellian relative speeds, scattering-rate utilities, and curated literature \emph{summary} constraints for regression tests and exploratory scans. On the rotation-curve side, we implement bounded non-linear least squares fits of NFW and Burkert halo models to SPARC baryonic decompositions, with optional mass-to-light priors and information-criterion summaries (AIC/BIC). For the demonstration dataset, we process 191 \texttt{rotmod} galaxies (LTG+ETG bundles) and fit both NFW and Burkert models (382 total fits). We find that Burkert is preferred by $\Delta \mathrm{BIC} > 0$ for $65.4\%$ of galaxies, with ``strong'' preference ($\Delta \mathrm{BIC}>6$) in $32.5\%$ of galaxies;

astro-ph.IM↗

arXiv2025

FANoS-v2: Feedback-Controlled Momentum with Thermostat Damping for Lightweight Neural Optimization

\FANOS{} is a PyTorch optimizer that augments RMS-preconditioned momentum with a scalar feedback controller over update energy. The public reference implementation stores momentum in parameter-update units, applies a non-negative thermostat damping coefficient, supports diagonal, factored, and raw-gradient preconditioning, and exposes diagnostics intended for stability audits. This study gives a complete mathematical specification of the released optimizer, including the exact parameter-unit update, the study-equation physical update mode, bounded log-ratio thermostat control, adaptive preconditioner softening, warmup guardrails, and the experimental \Fast{} profile. We report the v0.2 evidence: five-seed reduced-sample MNIST, Fashion-MNIST, and CIFAR-10 experiments show mean top-1 gains of 0.889, 2.197, and 2.666 percentage points over AdamW for \Fast{}, but with 49.8\%, 61.6\%, and 56.8\% higher wall-clock time. Preliminary scientific, PINN, and EEG smoke tests are mixed and are treated as hypothesis-generating only. The evidence supports \FANOS{} as an alpha-stage research optimizer with a reproducible lightweight-vision signal and an explicit runtime bottleneck.

cs.LG↗

arXiv2025

A Velocity Coupled Radial Acceleration Ansatz for Disk-Galaxy Rotation Curves: Fits to SPARC, Bayesian Inference, and Parameter Identifiability

Observed rotation curves of disk galaxies remain a sharp empirical probe of the relationship between baryons and dynamics. We study a minimal, explicitly \emph{phenomenological} alternative to standard halo parameterizations: an additional inward \emph{radial} acceleration proportional to the local \emph{tangential} speed, $a_{\vca}(r)=\gamma(r)\,v(r)$, with a saturating coupling $\gamma(r)=\vinf/(r+\rzero)$. Combining this ansatz with the circular-motion condition yields a quadratic equation for $v(r)$ with a closed-form physical branch.We fit this ``velocity-coupled acceleration'' (\vca) model to $N_\mathrm{gal}=171$ rotation curves from the SPARC sample using the published baryonic decompositions (gas, disk, bulge), and we compare to two commonly used two-parameter halo models (NFW and Burkert) using an identical optimization pipeline and error model.For a fiducial systematic error floor $\sigma_0=5~\mathrm{km\,s^{-1}}$, the \vca model is typically competitive with an NFW halo and performs comparably (though not uniformly better) than a Burkert halo in information-criterion comparisons.We further perform MCMC inference for \vca parameters, quantify posterior predictive coverage, and show that parameter posteriors exhibit a strong $\vinf$--$\rzero$ degeneracy for many galaxies; only $47/171$ galaxies yield well-identified posteriors under simple width-based criteria. We also perform a simple radial holdout cross-validation (outer 30\% of radii) and find predictive RMSE comparable to NFW and Burkert under this protocol.Finally..

astro-ph.GA↗