arXiv subjects

Hong-Yi Wang

Publications and source records attributed to Hong-Yi Wang.

13 recordsLinked to original sources

arXiv2026

Width-Independent Compressibility of Deep Neural Networks

It has long been known that well-trained neural networks can be compressed very strongly without affecting their performance, an important phenomenon that remains poorly understood. We prove a uniform compressibility theorem for deep multilayer perceptrons with analytic activations. For a deep, wide fixed teacher network, there exists a narrow (same depth) network that approximately represents the same function as the original. The reachable compressed width is strikingly independent of the original width, but is $O((\log(1/\varepsilon))^{d_{in}})$, where $\varepsilon$ is the error budget and $d_{in}$ is the effective input dimension. Our construction involves a novel derivative-matching technique which is aware of the low-dimensional input, and a layer-wise reweighting that preserves the input-output mapping.

cs.LG↗

arXiv2026

A semi-classical study of muon-enhanced proton-boron-11 fusion

A recent theoretical study has suggested that muons can enhance proton-boron-11 (p-$^{11}$B) reaction cross-section by several orders of magnitude in the low-energy regime. In this work, we investigate this reaction process using a semi-classical treatment, that is, a muon and a proton first form a muonic hydrogen atom p$μ$, which subsequently collides with a $^{11}$B nucleus. During the collision, the p$μ$ atom approaches the $^{11}$B nucleus and is then repelled by the Coulomb repulsive potential. At the distance of closest approach between the proton and the $^{11}$B nuclei in this classical scattering process -- namely, the classical turning point -- quantum tunneling through the Coulomb barrier can occur, allowing the proton to penetrate into the range of the nuclear force of the $^{11}$B and trigger the fusion reaction. We determine the turning point statistically by using the classical trajectory Monte Carlo method, where the initial phase-space distributions of the proton and muon are sampled from the ground-state microcanonical distribution. Our results show that, compared with the bare-nucleus case, the reaction cross-section is enhanced by several orders of magnitude in the low-energy region. A comparison with the static charge-shielding treatment reveals certain differences; however, both approaches demonstrate that the catalytic effect of the muon can significantly enhance the low-energy p-$^{11}$B reaction cross-section.

nucl-th↗

arXiv2026

A novel approach to proton-boron-11 fusion

Proton-boron-11 (p-$^{11}$B) fusion is a highly attractive aneutronic pathway for clean energy production, offering abundant fuel, negligible neutron activation, and the potential for direct energy conversion of charged $α$ particles. However, its practical implementation is severely hindered by the extremely high Coulomb barrier, necessitating ignition temperatures far beyond those of conventional deuterium-tritium reactions. In this work, we propose a novel approach to enhance the low-energy fusion cross-section by introducing a negative muon ($μ$). Instead of relying on the thermal equilibrium formation of a muonic molecule, we investigate a kinetic scenario in which a muonic hydrogen atom (p$μ$) is formed first and subsequently bombarded with a $^{11}$B nucleus. We quantitatively characterize the dynamic screening of the proton's Coulomb field by the tightly bound $μ$ cloud, the resulting modified Coulomb potential substantially lowers the effective barrier at intermediate separations. We also evaluate the penetrability, reaction cross-section, and reactivity of the p$μ$-$^{11}$B system, the results indicate that the inclusion of $μ$ enhances the tunneling probability by several orders of magnitude at incident energies below 100~keV, thereby significantly reducing the threshold for the nuclear reaction. This mechanism offers a promising alternative perspective for catalyzing p-$^{11}$B fusion, and also suggests a potential ignition pathway.

nucl-th↗

arXiv2026

A universal compression theory for lottery ticket hypothesis and neural scaling laws

When training large-scale models, the performance typically scales with the number of parameters and the dataset size according to a slow power law. A fundamental theoretical and practical question is whether comparable performance can be achieved with significantly smaller models and substantially less data. In this work, we provide a positive and constructive answer. We prove that a generic permutation-invariant function of $d$ objects can be asymptotically compressed into a function of $\operatorname{polylog} d$ objects with vanishing error, which is proved to be the optimal compression rate. This theorem yields two key implications: (Ia) a large neural network can be compressed to polylogarithmic width while preserving its learning dynamics; (Ib) a large dataset can be compressed to polylogarithmic size while leaving the loss landscape of the corresponding model unchanged. Implication (Ia) directly establishes a proof of the dynamical lottery ticket hypothesis, which states that any ordinary network can be strongly compressed such that the learning dynamics and result remain unchanged. (Ib) shows that a neural scaling law of the form $L\sim d^{-α}$ can be boosted to an arbitrarily fast power law decay, and ultimately to $\exp(-α' \sqrt[m]{d})$.

stat.ML↗

arXiv2026

Measuring unconventional causal structures in monitored dynamics

Causality underpins all logical reasoning. However, the causal structure in quantum processes can be far from intuitive, often differing from its classical counterpart in relativity, which is defined by the light cone. In particular, in systems with measurement and post-selection, causal influence can occur between spacelike separated regions. In this work, we study the causal structure and emergent "arrow of time" in monitored quantum dynamics, particularly their dependence on initial and final states. We propose a new measure, the cross-entropy quantum causal influence, to quantify the extent of causal influence, whose simulation demonstrates exotic causal structures, such as inverted light cones. This quantity can be measured in current quantum computing platforms. Additionally, we provide an analytical understanding of the relation between time arrow and entropy by studying two types of models that are analytically tractable: a quantum Brownian evolution model and a dual-unitary circuit model.

quant-ph↗

arXiv2026

Revisiting p-$^{11}$B Fusion: Updated Cross-sections, Reactivity, and Energy Balance

Recent experimental progress has substantially improved the available cross-section data for the p-$^{11}$B fusion reaction, particularly in energy regions that previously lacked direct measurements. In this study, we develop a high-precision analytical parameterization of the p-$^{11}$B reaction cross-section over the 0--10 MeV energy range, incorporating the new experimental data into a continuous and numerically efficient representation. Using this parameterization, we evaluate the thermonuclear reactivity of the p-$^{11}$B reaction and examine the effects of the dominant resonance at 0.6 MeV and a newly observed resonance around 4.7 MeV. Furthermore, we assess the energy balance by analyzing the fusion power density and the electron bremsstrahlung power density. Our results indicate that p-$^{11}$B fusion is not precluded by bremsstrahlung constraints when contemporary cross-section data and self-consistent thermal modeling are employed.

nucl-th↗

arXiv2025

Measurement induced scrambling and emergent symmetries in random circuits

Quantum entanglement is affected by unitary evolution, which spreads the entanglement through the whole system, and also by measurements, which usually tends to disentangle subsystems from the rest. Their competition has been known to result in the measurement-induced phase transition. But more intriguingly, measurement alone has the ability to drive a system into different entanglement phases. In this work, we map the entanglement evolution under unitaries and/or measurements into a classical spin problem. This framework is used to understand a myriad of random circuit models analytically, including measurement-induced and measurement-only transitions. Regarding many-body joint measurements, a lower bound of measurement range that is necessary for a global scrambled phase is derived. Moreover, emergent continuous symmetries (U(1) or SU(2)) are discovered in some random measurement models in the large-$d$ (qudit dimension) limit. The emergent continuous symmetry allows a variety of spin dynamics phenomena to find their counterparts in random circuit models.

quant-ph↗

arXiv2025

Relieving the post-selection problem by quantum singular value transformation

Quantum measurement is a fundamental yet experimentally challenging ingredient of quantum information processing. Many recent studies on quantum dynamics focus on expectation values of nonlinear observables; however, their experimental measurement is hindered by the post-selection problem -- namely, the substantial overhead caused by uncontrollable measurement outcomes. In this work, we propose a post-selection--free experimental strategy based on a fully quantum approach. The key idea is to deterministically simulate the post-selected quantum states by applying quantum singular value transformation (QSVT) algorithms. For pure initial state post-selection, our method is a generalization of fixed-point amplitude amplification to arbitrary projective measurements, achieving an optimal quadratic speedup. We further extend this framework to mixed initial state post-selection by applying linear amplitude amplification via QSVT, which significantly enhances the measurement success probability. However, a deterministic quantum algorithm for preparing the post-selected mixed state is generally impossible because of information-theoretic constraints imposed by quantum coding theory. Additionally, we introduce a pseudoinverse decoder for measurement-induced quantum teleportation. This decoder possesses the novel property that, when conditioned on a successful flag measurement, the decoding is nearly perfect even in cases where channel decoders are information-theoretically impossible. Overall, our work establishes a powerful approach for measuring novel quantum dynamical phenomena and presents quantum algorithms as a new perspective for understanding quantum dynamics and quantum chaos.

quant-ph↗

arXiv2024

Fragile non-Bloch spectrum and unconventional Green's function

In non-Hermitian systems, it is a counterintuitive feature of the non-Hermitian skin effect (NHSE) that the energy spectrum and eigenstates can be totally different under open or periodic boundary conditions, suggesting that non-Hermitian spectra can be extremely sensitive to non-local perturbations. Here, we show that a wide range of non-Hermitian models with NHSE can even be highly sensitive to local perturbation under open boundary conditions. The spectrum of these models is so fragile that it can be significantly modified by adding only exponentially small perturbations on boundaries. Intriguingly, we show that such fragile spectra are quantified by the Green's function exhibiting unconventional V-shape asymptotic behaviors. Accordingly, bi-directional exponential amplification can be observed. As an interesting consequence, we find a real-to-complex transition of the bulk spectrum induced by exponentially small boundary perturbations. Finally, we reveal a hierarchy of the asymptotic behaviors of non-Hermitian Green's functions, which restricts the frequency range for the presence of unconventional Green's functions.

quant-ph↗

arXiv2024

Amoeba Formulation of Non-Bloch Band Theory in Arbitrary Dimensions

The non-Hermitian skin effect dramatically reshapes the energy bands of non-Hermitian systems, meaning that the usual Bloch band theory is fundamentally inadequate as their characterization. The non-Bloch band theory, in which the concept of Brillouin zone is generalized, has been widely applied to investigate non-Hermitian systems in one spatial dimension. However, its generalization to higher dimensions has been challenging. Here, we develop a formulation of the non-Hermitian skin effect and non-Bloch band theory in arbitrary spatial dimensions, which is based on a natural geometrical object known as the amoeba. Our theory provides a general framework for studying non-Hermitian bands beyond one dimension. Key quantities of non-Hermitian bands, including the energy spectrum, eigenstates profiles, and the generalized Brillouin zone, can be efficiently obtained from this approach.

cond-mat.mes-hall↗

arXiv2024

Geometric Origin of Non-Bloch PT Symmetry Breaking

The parity-time (PT) symmetry of a non-Hermitian Hamiltonian leads to real (complex) energy spectrum when the non-Hermiticity is below (above) a threshold. Recently, it has been demonstrated that the non-Hermitian skin effect generates a new type of PT symmetry, dubbed the non-Bloch PT symmetry, featuring unique properties such as high sensitivity to the boundary condition. Despite its relevance to a wide range of non-Hermitian lattice systems, a general theory is still lacking for this generic phenomenon even in one spatial dimension. Here, we uncover the geometric mechanism of non-Bloch PT symmetry and its breaking. We find that non-Bloch PT symmetry breaking occurs by the formation of cusps in the generalized Brillouin zone (GBZ). Based on this geometric understanding, we propose an exact formula that efficiently determines the breaking threshold. Moreover, we predict a new type of spectral singularities associated with the symmetry breaking, dubbed non-Bloch van Hove singularity, whose physical mechanism fundamentally differs from their Hermitian counterparts. This singularity is experimentally observable in linear responses.

quant-ph↗

arXiv2022

Row-wise Accelerator for Vision Transformer

Following the success of the natural language processing, the transformer for vision applications has attracted significant attention in recent years due to its excellent performance. However, existing deep learning hardware accelerators for vision cannot execute this structure efficiently due to significant model architecture differences. As a result, this paper proposes the hardware accelerator for vision transformers with row-wise scheduling, which decomposes major operations in vision transformers as a single dot product primitive for a unified and efficient execution. Furthermore, by sharing weights in columns, we can reuse the data and reduce the usage of memory. The implementation with TSMC 40nm CMOS technology only requires 262K gate count and 149KB SRAM buffer for 403.2 GOPS throughput at 600MHz clock frequency.

cs.AR↗

arXiv2021

Non-Bloch PT symmetry breaking: Universal threshold and dimensional surprise

In the presence of non-Hermitian skin effect, non-Hermitian lattices generally have complex-valued eigenenergies under periodic boundary condition, but they can have non-Bloch PT symmetry and therefore completely real eigenenergies under open boundary condition. This novel PT symmetry and its breaking have been experimentally observed in one dimension. Here, we find that non-Bloch PT symmetry in two and higher dimensions exhibits drastically different behaviors compared to its one-dimensional counterpart. Whereas Bloch PT breaking and one-dimensional non-Bloch PT breaking generally have nonzero thresholds in the large-size limit, the threshold of two and higher-dimensional non-Bloch PT breaking universally approaches zero as the system size increases. A product measure, namely the product of bare non-Hermiticity and system size, is introduced to quantify the PT breaking tendency. This product being small is required for the perturbation theory to be valid, thus its growth with system size causes the breakdown of perturbation theory, which underlies the universal threshold. That the universal behaviors emerge only in two and higher dimensions indicates an unexpected interplay among PT symmetry, non-Hermitian skin effect, and spatial dimensionality. Our predictions can be confirmed on experimentally accessible platforms.

cond-mat.mes-hall↗