Searcharxiv⌕ Search

arXiv subjects

Takeshi Matsumura

Publications and source records attributed to Takeshi Matsumura.

At least 19 recordsLinked to original sources

Neural Network Verification for Deep Joint Source-Channel Coding

Deep joint source-channel coding (DeepJSCC) transmits data end-to-end over wireless channels using a neural encoder-decoder, but reconstruction quality can degrade sharply under adversarial perturbations and channel disturbances; no method formally bounds this degradation for DeepJSCC. We present the first bound-propagation framework for verifying DeepJSCC's decoder, bounding worst-case reconstruction error over a given wireless channel's noise region. Current deep neural network (DNN) verifiers do not support three DeepJSCC decoder components: parametric rectified linear activations (PReLU), transposed convolutions, and Rayleigh fading. We extend state-of-the-art techniques for optimization of linear relaxation in DNN verification for PReLU, replace the transposed convolution with its restricted upsample-then-convolution form, and formulate Rayleigh fading as a structural perturbation prepended directly into the decoder, thereby reducing the dimensionality of the verification problem. We also instantiate Lipschitz-regularized global robustness training, denoted GloRo, improving global robustness and enabling tight certification of DeepJSCC models for the first time. On DeepJSCC model for image transmission, this global robustness training procedure combined with structural encoding lowers the median certified bound by up to 41% and certifies about ten times more safe cases (192 against 19) than GloRo with interval encoding at a 10-degree error in channel estimation. Over-the-air validation with an orthogonal frequency-division multiplexing (OFDM) implementation on software-defined radio devices confirm the certificate holds on real hardware, with a worst observed error on radio link at 0.082 against a certified bound of 0.128.

cs.SE↗

Implementation of Zero-shot Semantic Communication on Software Defined Radio

Semantic communication has recently gained traction for its ability to reduce the amount of data transmitted over a communication link by transmitting a task-oriented representation instead of the raw source. Zero-shot semantic communication sends a general embedding from a vision-language model (VLM), so the same transmitter can serve new classification tasks without retraining. Most evidence for this advantage, however, comes from numerical simulation. We implement zero-shot semantic communication on a software-defined radio platform: a Raspberry Pi drives a pair of Analog Devices Active Learning Module (ADALM)-Pluto transceivers, with an image encoder at the transmitter and a text encoder at the receiver, and determines the zero-shot classification results via cosine similarity. We compare two VLMs, CLIP and MobileCLIP, across various channel conditions, i.e., different signal-to-noise ratios (SNRs). We validate that the semantic link spends 9x fewer channel uses per image than a JPEG plus 16-ary quadrature amplitude modulation baseline and still reaches 82% accuracy on CIFAR-10 at 22.3 dB, where the baseline scores 0%. On the traffic sign recognition dataset (TSRD), MobileCLIP correctly classifies 98.3% of unseen images at the same SNR. Offloading the image encoder to a neural processing unit reduces encoding to 13.4 ms per image, 49x faster than a Raspberry Pi 4 CPU, placing the transmitter within a real-time budget. Our implementation is publicly available at https://github.com/thanhlexyz/zsscsdr.

cs.NI↗

Helical-to-Fan Transitions under Magnetic Fields in the Noncentrosymmetric Tetragonal Magnet EuRhGe$_3$

The magnetic structure of EuRhGe$_3$, a noncentrosymmetric body-centered tetragonal magnet with the space group $I4mm$, has been investigated by resonant X-ray diffraction. Below $T_{\text{N}}=12$ K, EuRhGe$_3$ undergoes a helical magnetic ordering with an incommensurate propagation vector $q=(0, 0, 0.809)$, in which the magnetic moments lie in the $ab$ plane and rotate by a constant turn angle of $145.8^{\circ}$ between adjacent layers. When a magnetic field is applied along the $a$ axis at 2 K, a second-harmonic $2q$ peak develops, indicating that the circular helix is gradually distorted into a helimagnetic soliton-lattice state, which eventually undergoes a lock-in transition to the commensurate structure with $q=0.8$ at 3.8 T. Above the subsequent phase boundary at 5 T, the helicity is lost, and a spin-flop $xyz$-fan (elliptic conical) state is realized, in which the moments oscillate predominantly along the $b$ axis but are accompanied by a small $c$-axis component. At higher fields, the system enters a conventional planar $xy$-fan phase without a $c$-axis component. EuRhGe$_3$ provides a prototypical example of a helimagnet that exhibits a full sequence of field-induced structures, evolving from a circular helix to a spin-flop $xyz$-fan (elliptic conical), and finally to a planar $xy$-fan structure, which has been theoretically predicted.

cond-mat.str-el↗

Formal Verification for Deep Learning-based Power Control in Massive MIMO

Deep learning is a promising approach to optimize wireless communication by simplifying the search for near-optimal solutions. Prior studies on deep learning-based wireless communication optimization have explored supervised learning approaches that map raw user information, such as location or channel state information, to optimal power allocation vectors. While this approach demonstrates competitive performance, it is susceptible to adversarial attacks via input perturbations. Current defense mechanisms primarily rely on empirical methods, which do not provide formal guarantees of robustness. We fill this gap by proposing a formal verification framework to evaluate the robustness of deep learning-based power allocation in multi-cell massive multiple-input multiple-output (MIMO) systems against a wide range of potential adversarial input manipulations. To the best of our knowledge, this is the first attempt to formally verify deep neural networks in a regression setting with non-linear output constraints. We model the adversary's capabilities using hyper-rectangle constraints on their perturbation, adopt the abstraction-based bound-propagation technique (DeepPoly) to bound the interval of potential allocated powers, and formulate the minimum performance requirements as a constrained program for numerical feasibility analysis. Evaluation on publicly available datasets for power allocation in multi-cell massive MIMO indicates that a well-trained model can guarantee the local robustness under location perturbation by +-1m while retaining a maximum 1% optimality gap.

cs.NI↗

Collinear Magnetic Structure in the Diamond Network Magnet EuTi$_2$Al$_{20}$

The magnetic structure of EuTi$_2$Al$_{20}$, in which magnetic Eu$^{2+}$ ions form a diamond network, was investigated using neutron and resonant X-ray diffraction on powder and single-crystal samples. The propagation vector was determined to be $\textbf{\textit{q}}_{\rm m}=(1,0,0)$~r.l.u. from these diffraction measurements. All possible magnetic structures in the space group $Fd\bar{3}m$ with this propagation vector were examined using the irreducible representation method and magnetic space group analysis. This magnetic structure was identified as a collinear antiferromagnetic structure with the magnetic space group $P_Inna$ (\#52.320) or $P_Inn2$ (\#34.164) under zero magnetic field. In these magnetic structure, frustration arises from competing magnetic interactions on the diamond network. These findings provide a concrete experimental reference for assessing the role of competing interactions in diamond-network magnets and motivate further studies of interaction-driven quantum states.

cond-mat.str-el↗

Verifying DNN-based Semantic Communication Against Generative Adversarial Noise

Safety-critical applications like autonomous vehicles and industrial IoT are adopting semantic communication (SemCom) systems using deep neural networks to reduce bandwidth and increase transmission speed by transmitting only task-relevant semantic features. However, adversarial attacks against these DNN-based SemCom systems can cause catastrophic failures by manipulating transmitted semantic features. Existing defense mechanisms rely on empirical approaches provide no formal guarantees against the full spectrum of adversarial perturbations. We present VSCAN, a neural network verification framework that provides mathematical robustness guarantees by formulating adversarial noise generation as mixed integer programming and verifying end-to-end properties across multiple interconnected networks (encoder, decoder, and task model). Our key insight is that realistic adversarial constraints (power limitations and statistical undetectability) can be encoded as logical formulae to enable efficient verification using state-of-the-art DNN verifiers. Our evaluation on 600 verification properties characterizing various attacker's capabilities shows VSCAN matches attack methods in finding vulnerabilities while providing formal robustness guarantees for 44% of properties -- a significant achievement given the complexity of multi-network verification. Moreover, we reveal a fundamental security-efficiency tradeoff: compact 16-dimensional latent spaces achieve 50% verified robustness compared to 64-dimensional spaces.

cs.LO↗

Conical Magnetic Structure and Atomic Displacements in Chiral Helimagnet Yb(Ni,Cu)$_3$Al$_9$ in Magnetic Fields along the Helical $c$ Axis

We investigated the conical magnetic state of a uniaxial chiral helimagnet Yb(Ni$_{1-x}$Cu$_x$)$_3$Al$_9$ induced in magnetic fields applied along the $c$ axis, which coincides with the helical axis at zero field. Using resonant X-ray diffraction, we clearly observed the disappearance of magnetic satellite peaks, corresponding to the transition from the conical to the field-induced ferromagnetic state. The critical fields were determined to be 4 T for $x=0$ and 7 T for $x= 0.05$, which were hardly discernible in the magnetization curves. We also found that atomic displacements with the same propagation vector emerge simultaneously with the onset of the conical order. The transition temperature $T_{\text{N}}$ and the critical fields for $H \parallel c$ ($H_{\text{c}}^{z}$) and $H\perp c$ ($H_{\text{c}}^{x}$) are discussed on the basis of a mean-field calculation for a simple $q=1$ model of the magnetic structure. We propose that $T_{\text{N}}$ and $H_{\text{c}}^{z}$ primarily reflect the dominant intralayer exchange interactions within the honeycomb Yb-layer, whereas $H_{\text{c}}^{x}$ is governed by the much weaker interlayer coupling.

cond-mat.str-el↗

Acoustic phonon softening and lattice instability driven by on-site $f$-$d$ hybridization in CeCoSi

Soft phonon modes in tetragonal CeCoSi, which undergoes a structural transition at $T_0=12$ K followed by antiferromagnetic order at $T_{\text{N}}=9.5$ K, have been investigated using high-resolution inelastic x-ray scattering. Pronounced softening was detected in the transverse acoustic modes corresponding to the $(yz+zx)$-type monoclinic distortion, consistent with the experimentally determined triclinic structure. Remarkably, the softening persists up to the zone boundary along (0, 0, $q$), indicating a short correlation length of the lattice instability. This instability, characterized by a Curie-type strain susceptibility, is interpreted as a consequence of the on-site $4f$-$5d$ hybridization, which is intrinsic to this crystal structure due to the lack of inversion symmetry at the two Ce sites.

cond-mat.str-el↗

Single helicity of the triple-$q$ triangular skyrmion lattice state in cubic chiral helimagnet EuPtSi

We investigated the magnetic helicity of the triple-$q$ magnetic structure of the triangular skyrmion lattice in the ``A-phase" of EuPtSi for a magnetic field along the [111] axis by resonant x-ray diffraction using a circularly polarized beam. We show that all three Fourier components of the triple-$q$ structure are perpendicular to the respective $q$ vectors and have the same helicity. They are connected by the rotation operations about the [111] axis. The helicity is the same as that of the single-$q$ helimagnetic phase at low fields, suggesting that the antisymmetric exchange interaction inherent in the chiral structure supports the formation of the triangular skyrmion lattice. We also observe that the helical plane in the helimagnetic phase is tilted to the magnetic field to form a conical structure before the first-order transition to the skyrmion lattice phase.

cond-mat.str-el↗

Discovery of antiferromagnetic chiral helical ordered state in trigonal GdNi$_3$Ga$_9$

We have performed magnetic susceptibility, magnetization, and specific heat measurements on a chiral magnet GdNi$_3$Ga$_9$, belonging to the trigonal space group $R32$ (\#155). A magnetic phase transition takes place at $T_{\rm N}$ = 19.5 K. By applying a magnetic field along the $a$ axis at 2 K, the magnetization curve exhibits two jumps at $\sim$ 3 kOe and = 45 kOe. To determine the magnetic structure, we performed a resonant X-ray diffraction experiment by utilizing a circularly polarized beam. It is shown that a long-period antiferromagnetic (AFM) helical order is realized at zero field. The Gd spins in the honeycomb layer are coupled in an antiferromagnetic manner in the $c$ plane and rotate with a propagation vector $q$ = (0, 0, 1.485). The period of the helix is 66.7 unit cells ($\sim 180$~nm). In magnetic fields above 3~kOe applied perpendicular to the helical $c$ axis, the AFM helical order changes to an AFM order with $q$ = (0, 0, 1.5).

cond-mat.str-el↗

Distorted triangular skyrmion lattice in a noncentrosymmetric tetragonal magnet

Magnetic skyrmions are particle-like spin-swirling objects ubiquitously realized in magnets. They are topologically stable chiral kinks composed of multiple modulation waves of spiral spin structures, where the helicity of each spiral is usually selected by antisymmetric exchange interactions in noncentrosymmetric crystals. We report an experimental observation of a distorted triangular lattice of skyrmions in the polar tetragonal magnet EuNiGe$_3$, reflecting a strong coupling with the lattice. Moreover, through resonant x-ray diffraction, we find that the magnetic helicity of the original spiral at zero field is reversed when the skyrmion lattice is formed in a magnetic field. This means that the energy gain provided by the skyrmion lattice formation is larger than the antisymmetric exchange interaction. Our findings will lead us to a further understanding of emergent magnetic states.

cond-mat.str-el↗

Helicity Selection of the Cycloidal Order in Noncentrosymmetric EuIrGe$_3$

The magnetic helicities of the cycloidal ordering in EuIrGe$_3$, with a noncentrosymmetric tetragonal structure, have been studied by circularly polarized resonant X-ray diffraction. It is shown that the helicity of each cycloidal domain is uniquely determined and satisfies the symmetry relations of the $C_{4v}$ point group of the crystal structure. The result shows that the cycloidal helicity is determined by the Dzyaloshinskii-Moriya type antisymmetric exchange interaction. The domain selection and the phase transition by the external magnetic field along [100] and [110] have also been studied. It is shown that the cycloidal plane prefers to be perpendicular to the field and the transverse conical state is realized.

cond-mat.str-el↗

Crystal field excitation in the chiral helimagnet YbNi$_3$Al$_9$

Crystal field level scheme of a uniaxial chiral helimagnet YbNi$_3$Al$_9$, exhibiting a chiral magnetic soliton lattice state by Cu substitution for Ni, has been determined by inelastic neutron scattering. The ground and the first excited doublets are separated by 44 K and are simply expressed as $α|\pm 7/2\rangle + β|\mp 5/2\rangle$ with $α$ and $β$ nearly equal to $\pm 1/\sqrt{2}$. The easy axis of the crystal field anisotropy is the $c$ axis when the excited levels are populated at high temperatures and high magnetic fields. On the other hand, the magnetism at low temperatures and low magnetic fields, where only the ground doublet is populated, is described by an easy plane anisotropy which may be treated as an $S=1/2$ system with an anisotropic $g$-factor, $g_{xy}=3.02$ and $g_z=1.14$. An orbital dependent exchange interaction is also discussed to explain the temperature dependence of the magnetic susceptibility based on this level scheme.

cond-mat.str-el↗

Cycloidal Magnetic Ordering in Noncentrosymmetric EuIrGe$_3$

Successive magnetic phase transitions at $T_{\text{N}}$=12.2 K, $T_{\text{N}}^{\;\prime}$=7.0 K, and $T_{\text{N}}^{\;*}$=5.0 K in EuIrGe$_3$, an intermetallic compound with a body centered tetragonal lattice belonging to a polar space group $I4mm$, has been investigated by neutron diffraction and resonant X-ray diffraction. It is shown that EuIrGe$_3$ exhibits an incommensurate longitudinal sinusoidal order with $q\sim (0, 0, 0.792)$ and $m_{q} \parallel c\text{-axis}$ in the high temperature phase ($T_{\text{N}}^{\;\prime}< T < T_{\text{N}}$), which changes to a cycloidal order with $q=(δ', 0, 0.8)$ ($δ'\sim 0.017$) and $m_{q} \parallel ac\text{-plane}$ in the intermediate phase ($T_{\text{N}}^{\;*} < T < T_{\text{N}}^{\;\prime}$). In the low temperature phase ($T < T_{\text{N}}^{\;*}$), the cycloidal plane rotates by $45^{\circ}$ to have $q=(δ, δ, 0.8)$ ($δ\sim 0.012$). It is also pointed out that the X-ray scattering amplitude from odd-parity magnetic quadrupole due to the polar environment interfere with that from normal even-parity magnetic dipole in the magnetic ordered phase.

cond-mat.str-el↗

Structural Phase Transition and Possible Valence Instability of Ce$-4f$ Electron Induced by Pressure in CeCoSi

X-ray powder diffraction and electrical resistivity measurements were performed on the tetragonal compound CeCoSi under pressure to elucidate the phase boundary of the pressure-induced structural transition and the change in the 4$f$ electronic state. The temperature-pressure phase diagram has been determined from the shift of the Bragg peaks and from the anomaly in the resistivity. The critical pressure, $P_{\rm s}$ $\sim$ 4.9 GPa at 300 K, decreases to $P_{\rm s}$ $\sim$ 3.6 GPa at 10 K. The decrease of $P_{\rm s}$ is due not only to the decrease in volume of the unit cell but also to an anisotropic shrinkage by cooling. When crossing the boundary to the high-pressure phase, the resistivity shows a significant drop to exhibit a metallic temperature dependence. The results of this study strongly suggest that the structural phase transition can be ascribed to valence instability of Ce-$4f$ electron.

cond-mat.str-el↗

Structural Transition in the Hidden Ordered Phase of CeCoSi

We have performed X-ray diffraction experiments on a single crystalline CeCoSi to investigate the unresolved ordered phase below $T_0 \sim 12$ K. We have discovered that a triclinic lattice distortion takes place below $T_0$, which is further modified in the subsequent antiferromagnetic ordered phase. The structural domains can be selected by applying a magnetic field, indicating that some electronic ordering exists behind and affects the magnetic anisotropy in the hidden ordered phase below $T_0$. The transition at $T_0$, although the order parameter is still unknown, is associated with the maximum in the $c$-axis lattice parameter. In magnetic fields along $[1, 0, 0]$, the structural transition temperature, named as $T_{\text{s1}}$, deviates from $T_0$ and decreases with increasing the field, whereas $T_0$ increases. This shows that the hidden ordered phase without triclinic distortion exists between $T_{\text{s1}}$ and $T_0$. The results for $H \parallel [1, 1, 0]$ are also reported.

cond-mat.str-el↗

Chirality-Controlled Enantiopure Crystal Growth of a Transition Metal Monosilicide by a Floating Zone Method

We performed a crystal growth to obtain chirality-controlled enantiopure crystals using a laser-diode-heated floating zone (LDFZ) method with a composition-gradient feed rod. It has been argued that the crystal handedness of $T$Si ($T$ : transition metal) is fixed depending on $T$ in the case of the ones grown by the conventional methods. We found that right-handed single crystals of CoSi and MnSi were grown from the composition gradient feed rods that consist of FeSi--CoSi and FeSi--MnSi, respectively. The obtained CoSi and MnSi crystals inherit the chirality from the seed part of FeSi, which grows in a right-handed structure, and thus have the chirality opposite to that for the crystals in the literature. The LDFZ method with the feed rods with various combinations of $T$Si compounds enables a flexible control of the chirality of $T$Si and will be useful for clarifying the interplay between the crystalline chirality and chirality-induced physical responses.

cond-mat.mtrl-sci↗

Competition between helimagnetic and ferroquadrupolar orderings in a monoaxial chiral magnet DyNi$_3$Ga$_{9}$ studied by resonant x-ray diffraction

Successive phase transitions in a rare-earth monoaxial chiral magnet DyNi$_3$Ga$_{9}$ have been investigated by resonant x-ray diffraction. Magnetic dipole and electric quadrupole degrees of freedom arising from the large angular moment of $J=15/2$, in combination with the symmetric and antisymmetric exchange interactions and the crystal field anisotropy, give rise to competing ordered phases. We show that the antiferromagnetically coupled Dy moments in the $ab$-plane form an incommensurate helimagnetic order with $q\sim(0, 0, 0.43)$ just below $T_{\text{N}}=10$ K, which further exhibits successive first-order transitions to the commensurate helimagnetic order with $q=(0,0,0.5)$ at $T_{\text{N}}^{\;\prime}=9.0$ K, and to the canted antiferromagnetic order with $q=(0,0,0)$ at $T_{\text{N}}^{\;\prime\prime}=8.5$ K, both with large coexistence regions. The relation of the magnetic helicity and the crystal chirality in DyNi$_3$Ga$_{9}$ is also uniquely determined. Splitting of the $(6,0,0)$ Bragg peak is observed below $T_{\text{N}}^{\;\prime\prime}$, reflecting the lattice distortion due to the ferroquadrupole order. In the canted antiferromagnetic phase, a spin-flop transition takes place at 5 K when the temperature is swept in a weak magnetic field. We discuss these transitions from the viewpoint of competing energies described above.

cond-mat.str-el↗