SearcharxivSearch

arXiv subjects

Arpan Mukherjee

Publications and source records attributed to Arpan Mukherjee.

At least 19 recordsLinked to original sources

Finite Spectral-Band Optimal Control of Acoustic Waves via Subwavelength Point-Like Resonant Actuators

We study finite-band optimal control of acoustic waves actuated by local clusters of subwavelength resonators. The acoustic problem reduces to a time-domain Foldy-Lax approximation capturing wave-structure interaction. Spectral analysis of the delayed transfer matrix isolates collective scattering resonances corresponding to weakly damped poles $s_\alpha^\epsilon=-\gamma_\alpha^\epsilon+i\omega_\alpha^\epsilon$ with radiation damping $\gamma_\alpha^\epsilon>0$. Projecting onto a finite band yields the coupled system $\ddot{a}+\Lambda a=C_\epsilon\eta$, $\ddot{\eta}+2\Gamma_\epsilon\dot{\eta}+K_\epsilon\eta=u$, where $a$, $\eta$, and $u$ are modal coefficients, microstructural states, and control. For a tracking functional $\mathcal{J}_{\mu}$ with regularization $\mu>0$, we prove existence and uniqueness of the optimal control and derive the adjoint system. Our main quantitative result is a resonant source-lifting estimate: if a source profile $\eta_r$ is spectrally concentrated in bands $I_\alpha$, the input $u_r=(\partial_t^2+2\Gamma_\epsilon\partial_t+K_\epsilon)\eta_r$ satisfies $\|u_r\|_{L^2(0,T)}^2 \le \sum_\alpha \left(\sup_{\nu\in I_\alpha} |(\omega_\alpha^\epsilon)^2+(\gamma_\alpha^\epsilon)^2-\nu^2 +2i\gamma_\alpha^\epsilon\nu|\right)^2 \|(\eta_r)_\alpha\|_{\mathcal{B}_T(I_\alpha)}^2$. This provides an upper bound for the optimal value function. At exact matching $\nu=\omega_\alpha^\epsilon$, the multiplier equals $2\gamma_\alpha^\epsilon\omega_\alpha^\epsilon+O((\gamma_\alpha^\epsilon)^3)$, showing clustering yields a finite resonant gain governed by the pole's real part. Finally, this attenuation enables finite-band stabilization under an explicit modal coupling condition, with a decay rate proportional to the cluster damping scale.

math.AP

Resonant Microstructures as Dirac-type Actuators for Acoustic Wave Control

We study interior control of the acoustic wave equation via effective point sources generated by a finite cluster of resonant perturbations (modeling acoustic subwavelength bubbles). At the abstract level, after localizing the whole-space dynamics to a large auxiliary observation domain, we consider a Dirichlet spectral formulation of the wave equation with finitely many point actuators located at prescribed interior points. Restricting to a finite spectral band of Dirichlet eigenfrequencies, we prove that, under a natural full-rank condition on the associated coupling matrix, arbitrary trajectories on the corresponding spectral subspace can be realized, with quantitative bounds on the control cost in terms of spectral-band geometry and actuator placement. We then show that these ideal actuators can be realized by clusters of small, high-contrast bubbles. Using a time-domain asymptotic expansion, the scattered field is represented as a superposition of retarded monopoles whose amplitudes satisfy a finite-dimensional delayed hyperbolic system. In the Laplace domain, this induces a transfer operator whose pole structure encodes the Minnaert resonance with a collective attenuation. We prove that the associated actuator map is ill-conditioned away from resonance, whereas, under a cluster-level transducer accessibility condition linking the incident fields to the dominant cluster channels, it admits a bounded right inverse on suitable Minnaert bands. Consequently, one obtains spectral tracking of the wave field with error $\mathcal{O}(\varepsilon^\gamma)$ as the bubble size $\varepsilon \to 0$. Keywords. Wave equation, Dirac actuators, Trajectory tracking control, Resonant perturbations, Kato's analytic perturbation, Perron-Frobenius spectrum, Minnaert resonances, Actuation map, Toeplitz matrix.

math.AP

Feedback Stabilization and Tracking for Heat Equations Using Thermo-Plasmonic Nanoparticles as Actuators

We propose a feedback strategy to track prescribed heat profiles using plasmonic nanoparticles as actuators. Starting from a thermo--plasmonic Maxwell--heat model, we use a time-domain discrete effective description in which the generated heat is approximated by a superposition of heat kernels centered at particle locations with amplitudes governed by a coupled Volterra system. We recast this dynamics as a heat equation on a bounded domain with finitely many point actuators and design a tracking feedback based on pointwise evaluations of $\mathcal A^{-1}y$, where $\mathcal A=I-A_0$ and $A_0$ is the Neumann diffusion operator. Working in the natural $V'$ setting with $V=D(\mathcal A)$, we prove exponential stabilization of the tracking error via distribution-actuator theory. For non-equilibrium reference profiles, we add a constant feedforward term and a low-mode fixed-point pre-compensation on $X_N$, ensuring exact steady matching on $X_N$ and an explicit bound on the residual tail mismatch.

math.AP

Functional-Analytic Justification of the Time-Domain Foldy-Lax Approximation for Dispersive Acoustic Media: A Feynman-Diagram Viewpoint

This work provides a rigorous functional-analytic justification for a time-domain Foldy-Lax framework that describes multiple acoustic scattering by a cluster of dispersive resonators (modeling gas-filled bubbles), explicitly incorporating dispersion via the Minnaert resonance. The model is formulated as a delayed-coupled hyperbolic system for bubble amplitude interactions. We combine time-domain integral equations, Laplace transforms, and Hardy-Sobolev space techniques to analyze this system, establishing its unique solvability in anisotropic Hilbert spaces, with solutions expressed as convergent Neumann series of convolution operators. We derive geometric decay of truncation errors for resonant incident waves and quantify the contribution of $N$-th order multi-scattering, showing it scales with \(\varepsilon^{N(1-p)+1}\) (relating bubble radius \(\varepsilon\) and inter-bubble distance scaling as $\varepsilon^p$, $p<1$). This dominates the measurement errors, which are of order $\varepsilon^2$, thereby allowing us to capture fields generated by inter-bubble interactions of order $N<\frac{1}{1-p}$. This provides a quantitative relation between the spectra band width of the source field, the closeness distance between the bubbles and the order $N$ of the relevant interactions between the bubbles. Furthermore, a novel connection to Feynman diagrams maps multi-scattering paths to diagrammatic vertices and propagators, simplifying the interpretation of higher-order interactions and kinematic constraints. This framework advances accurate transient wave prediction in dispersive media, with implications for cavitation therapy, seismic imaging, and metamaterial engineering.

math.AP

Test-time Verification via Optimal Transport: Coverage, ROC, & Sub-optimality

While test-time scaling with verification has shown promise in improving the performance of large language models (LLMs), the role of the verifier and its imperfections remain underexplored. The effect of verification manifests through interactions of three quantities: (i) the generator's coverage, (ii) the verifier's region of convergence (ROC), and (iii) the sampling algorithm's sub-optimality. Though recent studies capture subsets of these factors, a unified framework quantifying the geometry of their interplay is missing. We frame verifiable test-time scaling as a transport problem. This characterizes the interaction of coverage, ROC, and sub-optimality, and uncovers that the sub-optimality--coverage curve exhibits three regimes. A transport regime -- where sub-optimality increases with coverage, a policy improvement regime -- where sub-optimality may decrease with coverage, depending on the verifier's ROC, and a saturation regime -- where sub-optimality plateaus, unaffected by coverage. We further propose and analyze two classes of sampling algorithms -- sequential and batched, and examine how their computational complexities shape these trade-offs. Empirical results with Qwen, Llama, and Gemma models corroborate our theoretical findings.

cs.AI

SharedRep-RLHF: A Shared Representation Approach to RLHF with Diverse Preferences

Uniform-reward reinforcement learning from human feedback (RLHF), which trains a single reward model to represent the preferences of all annotators, fails to capture the diversity of opinions across sub-populations, inadvertently favoring dominant groups. The state-of-the-art, MaxMin-RLHF, addresses this by learning group-specific reward models, and by optimizing for the group receiving the minimum reward, thereby promoting fairness. However, we identify that a key limitation of MaxMin-RLHF is its poor performance when the minimum-reward group is a minority. To mitigate this drawback, we introduce a novel framework, termed {\em SharedRep-RLHF}. At its core, SharedRep-RLHF learns and leverages {\em shared traits} in annotations among various groups, in contrast to learning separate reward models across groups. We first show that MaxMin-RLHF is provably suboptimal in learning shared traits, and then quantify the sample complexity of SharedRep-RLHF. Experiments across diverse natural language tasks showcase the effectiveness of SharedRep-RLHF compared to MaxMin-RLHF with a gain of up to 20% in win rate.

cs.LG

Preference-centric Bandits: Optimality of Mixtures and Regret-efficient Algorithms

The objective of canonical multi-armed bandits is to identify and repeatedly select an arm with the largest reward, often in the form of the expected value of the arm's probability distribution. Such a utilitarian perspective and focus on the probability models' first moments, however, is agnostic to the distributions' tail behavior and their implications for variability and risks in decision-making. This paper introduces a principled framework for shifting from expectation-based evaluation to an alternative reward formulation, termed a preference metric (PM). The PMs can place the desired emphasis on different reward realization and can encode a richer modeling of preferences that incorporate risk aversion, robustness, or other desired attitudes toward uncertainty. A fundamentally distinct observation in such a PM-centric perspective is that designing bandit algorithms will have a significantly different principle: as opposed to the reward-based models in which the optimal sampling policy converges to repeatedly sampling from the single best arm, in the PM-centric framework the optimal policy converges to selecting a mix of arms based on specific mixing weights. Designing such mixture policies departs from the principles for designing bandit algorithms in significant ways, primarily because of uncountable mixture possibilities. The paper formalizes the PM-centric framework and presents two algorithm classes (horizon-dependent and anytime) that learn and track mixtures in a regret-efficient fashion. These algorithms have two distinctions from their canonical counterparts: (i) they involve an estimation routine to form reliable estimates of optimal mixtures, and (ii) they are equipped with tracking mechanisms to navigate arm selection fractions to track the optimal mixtures. These algorithms' regret guarantees are investigated under various algebraic forms of the PMs.

stat.ML

Risk-sensitive Bandits: Arm Mixture Optimality and Regret-efficient Algorithms

This paper introduces a general framework for risk-sensitive bandits that integrates the notions of risk-sensitive objectives by adopting a rich class of distortion riskmetrics. The introduced framework subsumes the various existing risk-sensitive models. An important and hitherto unknown observation is that for a wide range of riskmetrics, the optimal bandit policy involves selecting a mixture of arms. This is in sharp contrast to the convention in the multi-arm bandit algorithms that there is generally a solitary arm that maximizes the utility, whether purely reward-centric or risk-sensitive. This creates a major departure from the principles for designing bandit algorithms since there are uncountable mixture possibilities. The contributions of the paper are as follows: (i) it formalizes a general framework for risk-sensitive bandits, (ii) identifies standard risk-sensitive bandit models for which solitary arm selections is not optimal, (iii) and designs regret-efficient algorithms whose sampling strategies can accurately track optimal arm mixtures (when mixture is optimal) or the solitary arms (when solitary is optimal). The algorithms are shown to achieve a regret that scales according to $O((\log T/T )^{\nu})$, where $T$ is the horizon, and $\nu>0$ is a riskmetric-specific constant.

stat.ML

Dispersive Effective Metasurface Model for Bubbly Media

We derive the effective transmission condition for a cluster of acoustic subwavelength resonators, modeled as small-scaled bubbles distributed not necessarily periodically along a smooth, bounded hypersurface, which need not be flat. The transmission condition specifies that the jump in the normal derivative of the acoustic field is proportional to its second time derivative, convoluted in time with a sinusoidal kernel. This kernel has a period determined by the common subwavelength resonance (specifically, the Minnaert resonance in this case). This dispersive transmission condition can also be interpreted as a Dirac-like surface potential that is convoluted in the time domain and spatially supported on the specified hypersurface. We highlight the following features: 1. High resonance regime: When the common resonance is large, the surface behaves as fully transparent, permitting complete transmission of the acoustic field. 2. Moderate resonance regime: For moderate resonance values, the surface acts as a screen with memory effects, capturing the dispersive behavior induced by the resonance. 3. Low resonance regime: When the common resonance is small, the surface functions as a partial reflective (or partial transmissive) screen with no memory effect.

math.AP

Effective Medium Theory for Heat Generation Using Plasmonics: A Parabolic Transmission Problem Driven by the Maxwell System

The excitation of plasmonic nanoparticles by incident electromagnetic waves at frequencies near their subwavelength resonances induces localized heat generation in the surrounding medium. We develop a mathematical framework to rigorously quantify this heat generation in systems of arbitrarily distributed nanoparticles. 1. For an arbitrary discrete distribution of M nanoparticles within a bounded domain, the effective heat distribution is described by a coupled system: Volterra-type integral equations for the heat conduction and a Foldy-Lax-type system governing the self consistent electric field intensities. These equations are parameterized by the particle geometries and the local electromagnetic field interactions. The effective heat generation is computed by solving these coupled systems, with the computational complexity scaling as M^2. 2. In the case M >> 1, under natural scaling regimes, the discrete system converges to a continuum model, yielding an effective parabolic equation for the heat distribution. The source term in this homogenized parabolic model is characterized by the solution of the homogenized Maxwells equations, incorporating an effective permittivity distribution derived from the Drude model under resonance conditions. Our analysis utilizes advanced tools in potential theory, asymptotic analysis and homogenization. By leveraging layer potential representations, we derive point-wise field approximations. The coupling between the Maxwell and heat equations is resolved by analyzing the spectral properties of the nanoparticles and their scaling limits. This framework reduces the problem to two mathematical challenges: a control problem for the effective parabolic system and an internal phase-less inverse problem for the Maxwell system, thus providing a unified approach to modeling heat generation in nanoparticle clusters.

math.AP

Combinatorial Multi-armed Bandits: Arm Selection via Group Testing

This paper considers the problem of combinatorial multi-armed bandits with semi-bandit feedback and a cardinality constraint on the super-arm size. Existing algorithms for solving this problem typically involve two key sub-routines: (1) a parameter estimation routine that sequentially estimates a set of base-arm parameters, and (2) a super-arm selection policy for selecting a subset of base arms deemed optimal based on these parameters. State-of-the-art algorithms assume access to an exact oracle for super-arm selection with unbounded computational power. At each instance, this oracle evaluates a list of score functions, the number of which grows as low as linearly and as high as exponentially with the number of arms. This can be prohibitive in the regime of a large number of arms. This paper introduces a novel realistic alternative to the perfect oracle. This algorithm uses a combination of group-testing for selecting the super arms and quantized Thompson sampling for parameter estimation. Under a general separability assumption on the reward function, the proposed algorithm reduces the complexity of the super-arm-selection oracle to be logarithmic in the number of base arms while achieving the same regret order as the state-of-the-art algorithms that use exact oracles. This translates to at least an exponential reduction in complexity compared to the oracle-based approaches.

cs.LG

Uncertainty-Informed Screening for Safer Solvents Used in the Synthesis of Perovskite via Language Models

The challenge of accurately predicting toxicity of industrial solvents used in perovskite synthesis is a necessary undertaking but is limited by a lack of a targeted and structured toxicity data. This paper presents a novel framework that combines an automated data extraction using language models, and an uncertainty-informed prediction model to fill data gaps and improve prediction confidence. First, we have utilized and compared two approaches to automatically extract relevant data from a corpus of scientific literature on solvents used in perovskite synthesis: smaller bidirectional language models like BERT and ELMo are used for their repeatability and deterministic outputs, while autoregressive large language model (LLM) such as GPT-3.5 is used to leverage its larger training corpus and better response generation. Our novel 'prompting and verification' technique integrated with an LLM aims at targeted extraction and refinement, thereby reducing hallucination and improving the quality of the extracted data using the LLM. Next, the extracted data is fed into our pre-trained multi-task binary classification deep learning to predict the ED nature of extracted solvents. We have used a Shannon entropy-based uncertainty quantification utilizing the class probabilities obtained from the classification model to quantify uncertainty and identify data gaps in our predictions. This approach leads to the curation of a structured dataset for solvents used in perovskite synthesis and their uncertainty-informed virtual toxicity assessment. Additionally, chord diagrams have been used to visualize solvent interactions and prioritize those with potential hazards, revealing that 70% of the solvent interactions were primarily associated with two specific perovskites.

physics.chem-ph

Time-Dependent Acoustic Waves Generated by Multiple Resonant Bubbles: Application to Acoustic Cavitation

We analyse the ultrasound waves reflected by multiple bubbles in the linearized time-dependent acoustic model. The generated time-dependent wave field is estimated close to the bubbles. The motivation of this study comes from the therapy modality using acoustic cavitation generated by injected bubbles into the region of interest. The goal is to create enough, but not too much, pressure in the region of interest to eradicate anomalies in that region. Here, we derive the dominant part of the generated acoustic field by a cluster of bubbles taking into account the (high) contrasts of their mass density and bulk as well as their general distribution in the given region. As consequences of these approximations, we highlight the following features: 1. If we use dimers (two close bubbles), or generally polymers, then we obtain a remarkable enhancement of the whole echo in the whole time. 2. If we distribute the bubbles every where in the region of interest,then we can derive the effective acoustic model which turns out to be a dispersive one. We show that, for a given desired pressure, we can tune the effective model to generate it.

math.AP

Dispersive Effective Model in the Time-Domain for Acoustic Waves Propagating in Bubbly Media

We derive the effective medium theory for the linearized time-domain acoustic waves propagating in a bubbly media. The analysis is done in the time-domain avoiding the need to use Fourier transformation. This allows considering general incident waves avoiding band limited ones as usually used in the literature. Most importantly, the outcome is as follows: 1. As the bubbles are resonating, with the unique subwavelength Minnaert resonance, the derived effective wave model is dispersive. Precisely, the effective acoustic model is an integro-differential one with a time-convolution term highlighting the resonance effect. 2. The periodicity in distributing the cluster of bubbles is not needed, contrary to the case of using traditional two-scale homogenization procedures. Precisely, given any $C^1$-smooth function $K$, we can distribute the bubbles so that locally the number of such bubbles is dictated by $K$. In addition to its dispersive character, the effective model is affected by the function $K$. Such freedom and generality is appreciable in different applied sciences including materials sciences and mathematical imaging.

math.AP

Optimal Best Arm Identification with Fixed Confidence in Restless Bandits

We study best arm identification in a restless multi-armed bandit setting with finitely many arms. The discrete-time data generated by each arm forms a homogeneous Markov chain taking values in a common, finite state space. The state transitions in each arm are captured by an ergodic transition probability matrix (TPM) that is a member of a single-parameter exponential family of TPMs. The real-valued parameters of the arm TPMs are unknown and belong to a given space. Given a function $f$ defined on the common state space of the arms, the goal is to identify the best arm -- the arm with the largest average value of $f$ evaluated under the arm's stationary distribution -- with the fewest number of samples, subject to an upper bound on the decision's error probability (i.e., the fixed-confidence regime). A lower bound on the growth rate of the expected stopping time is established in the asymptote of a vanishing error probability. Furthermore, a policy for best arm identification is proposed, and its expected stopping time is proved to have an asymptotic growth rate that matches the lower bound. It is demonstrated that tracking the long-term behavior of a certain Markov decision process and its state-action visitation proportions are the key ingredients in analyzing the converse and achievability bounds. It is shown that under every policy, the state-action visitation proportions satisfy a specific approximate flow conservation constraint and that these proportions match the optimal proportions dictated by the lower bound under any asymptotically optimal policy. The prior studies on best arm identification in restless bandits focus on independent observations from the arms, rested Markov arms, and restless Markov arms with known arm TPMs. In contrast, this work is the first to study best arm identification in restless bandits with unknown arm TPMs.

stat.ML

Improved Bound for Robust Causal Bandits with Linear Models

This paper investigates the robustness of causal bandits (CBs) in the face of temporal model fluctuations. This setting deviates from the existing literature's widely-adopted assumption of constant causal models. The focus is on causal systems with linear structural equation models (SEMs). The SEMs and the time-varying pre- and post-interventional statistical models are all unknown and subject to variations over time. The goal is to design a sequence of interventions that incur the smallest cumulative regret compared to an oracle aware of the entire causal model and its fluctuations. A robust CB algorithm is proposed, and its cumulative regret is analyzed by establishing both upper and lower bounds on the regret. It is shown that in a graph with maximum in-degree $d$, length of the largest causal path $L$, and an aggregate model deviation $C$, the regret is upper bounded by $\tilde{\mathcal{O}}(d^{L-\frac{1}{2}}(\sqrt{T} + C))$ and lower bounded by $Ω(d^{\frac{L}{2}-2}\max\{\sqrt{T}\; ,\; d^2C\})$. The proposed algorithm achieves nearly optimal $\tilde{\mathcal{O}}(\sqrt{T})$ regret when $C$ is $o(\sqrt{T})$, maintaining sub-linear regret for a broad range of $C$.

stat.ML

Robust Causal Bandits for Linear Models

Sequential design of experiments for optimizing a reward function in causal systems can be effectively modeled by the sequential design of interventions in causal bandits (CBs). In the existing literature on CBs, a critical assumption is that the causal models remain constant over time. However, this assumption does not necessarily hold in complex systems, which constantly undergo temporal model fluctuations. This paper addresses the robustness of CBs to such model fluctuations. The focus is on causal systems with linear structural equation models (SEMs). The SEMs and the time-varying pre- and post-interventional statistical models are all unknown. Cumulative regret is adopted as the design criteria, based on which the objective is to design a sequence of interventions that incur the smallest cumulative regret with respect to an oracle aware of the entire causal model and its fluctuations. First, it is established that the existing approaches fail to maintain regret sub-linearity with even a few instances of model deviation. Specifically, when the number of instances with model deviation is as few as $T^\frac{1}{2L}$, where $T$ is the time horizon and $L$ is the longest causal path in the graph, the existing algorithms will have linear regret in $T$. Next, a robust CB algorithm is designed, and its regret is analyzed, where upper and information-theoretic lower bounds on the regret are established. Specifically, in a graph with $N$ nodes and maximum degree $d$, under a general measure of model deviation $C$, the cumulative regret is upper bounded by $\tilde{\mathcal{O}}(d^{L-\frac{1}{2}}(\sqrt{NT} + NC))$ and lower bounded by $Ω(d^{\frac{L}{2}-2}\max\{\sqrt{T},d^2C\})$. Comparing these bounds establishes that the proposed algorithm achieves nearly optimal $\tilde{\mathcal{O}}(\sqrt{T})$ regret when $C$ is $o(\sqrt{T})$ and maintains sub-linear regret for a broader range of $C$.

stat.ML

Acoustic Cavitation using Resonating Micro-Bubbles. Analysis in the Time-Domain

We study the time-domain acoustic wave propagation in the presence of a micro-bubble. This micro-bubble is characterized by a mass density and bulk modulus which are both very small as compared to the ones of the background vacuum. The goal is to estimate the amount of pressure that is created very near (at a distance proportional to the radius of the bubble) to the bubble. We show that, at that small distance, the dominating field is reminiscent to the wave created by a point-like obstacle modeled formally by a Dirac-like heterogeneity with support at the location of the bubble and the contrast between the bubble and background material as the scattering coefficient. As a conclusion, we can tune the bubbles material properties so that the pressure near it reaches a desired amount. Such design might be useful in the purpose of acoustic cavitation where one needs enough, but not too much, pressure to eliminate unwanted anomalies. The mathematical analysis is done using time-domain integral equations and asymptotic analysis techniques. A well known feature here is that the contrasting scales between the bubble and the background generate resonances (mainly the Minnaert one) in the time-harmonic regime. Such critical scales, and the generated resonances, are also reflected in the time-domain estimation of the acoustic wave. In particular, reaching the desired amount of pressure near the location of the bubble is possible only with such resonating bubbles.

math.AP