Search

The advent of chiplet technology introduces cutting-edge opportunities for constructing highly heterogeneous platforms with specialized accelerators. However, the HPC community currently lacks expertise in hardware development, a gap that must be bridged to leverage these advancements. Additionally, technologies like chiplet is cutting-edge with limited educational resource available. This paper addresses potential hardware specialization direction in HPC and how to cultivate these skills among students and staff, emphasizing the importance of understanding and developing custom hardware (e.g., rapid prototyping and resource estimation). We have been mentoring graduate-level students and new staff in hardware designs in a hands-on manner, encouraging them to utilize modern open-source hardware tools for their designs, which facilitates the sharing of research ideas. Additionally, we provide a summary of theses tools as part of our approach to prototyping and mentoring., Comment: SC24 eduHPC 2 pages

This paper describes speech enhancement for realtime automatic speech recognition (ASR) in real environments. A standard approach to this task is to use neural beamforming that can work efficiently in an online manner. It estimates the masks of clean dry speech from a noisy echoic mixture spectrogram with a deep neural network (DNN) and then computes a enhancement filter used for beamforming. The performance of such a supervised approach, however, is drastically degraded under mismatched conditions. This calls for run-time adaptation of the DNN. Although the ground-truth speech spectrogram required for adaptation is not available at run time, blind dereverberation and separation methods such as weighted prediction error (WPE) and fast multichannel nonnegative matrix factorization (FastMNMF) can be used for generating pseudo groundtruth data from a mixture. Based on this idea, a prior work proposed a dual-process system based on a cascade of WPE and minimum variance distortionless response (MVDR) beamforming asynchronously fine-tuned by block-online FastMNMF. To integrate the dereverberation capability into neural beamforming and make it fine-tunable at run time, we propose to use weighted power minimization distortionless response (WPD) beamforming, a unified version of WPE and minimum power distortionless response (MPDR), whose joint dereverberation and denoising filter is estimated using a DNN. We evaluated the impact of run-time adaptation under various conditions with different numbers of speakers, reverberation times, and signal-to-noise ratios (SNRs)., Comment: Accepted to APSIPA2024

This paper describes sound event localization and detection (SELD) for spatial audio recordings captured by firstorder ambisonics (FOA) microphones. In this task, one may train a deep neural network (DNN) using FOA data annotated with the classes and directions of arrival (DOAs) of sound events. However, the performance of this approach is severely bounded by the amount of annotated data. To overcome this limitation, we propose a novel method of pretraining the feature extraction part of the DNN in a self-supervised manner. We use spatial audio-visual recordings abundantly available as virtual reality contents. Assuming that sound objects are concurrently observed by the FOA microphones and the omni-directional camera, we jointly train audio and visual encoders with contrastive learning such that the audio and visual embeddings of the same recording and DOA are made close. A key feature of our method is that the DOA-wise audio embeddings are jointly extracted from the raw audio data, while the DOA-wise visual embeddings are separately extracted from the local visual crops centered on the corresponding DOA. This encourages the latent features of the audio encoder to represent both the classes and DOAs of sound events. The experiment using the DCASE2022 Task 3 dataset of 20 hours shows non-annotated audio-visual recordings of 100 hours reduced the error score of SELD from 36.4 pts to 34.9 pts., Comment: Accepted to APSIPA2023

Atomic crystals with dislocations deform plastically at low stresses via dislocation glide. Whether dislocation glide occurs in macroscopic frictional granular media has remained unknown. The discrete element method is employed to simulate the structural and mechanical responses of a granular crystal with an edge dislocation. We find that dislocation glide occurs at low interparticle friction, resulting in significantly lower yield stresses than in dislocation-free crystals. Yield stress varies linearly with interparticle friction, attributed to both Peierls stress and frictional effect., Comment: 5 pages, 4 figures

This paper explores how magnetic fields affect the Casimir effect within the context of a simple quasi-1D interacting fermionic system. A novel phenomenon emerges, resulting from the interaction between external magnetic fields and boundary conditions, which alters the ground state in complex ways and leads to first-order phase transitions among various ground states, specifically inhomogeneous solutions with differing node counts. We calculate the Casimir force, which exhibits discontinuities (jumps) at the transition points between the different ground states., Comment: 6 pages, 5 figures

We describe the structure of projective indecomposable modules for the subalgebra consisting of the elements of degree 0 in the hyperalgebra of the $r$-th Frobenius kernel for the algebraic group ${\rm SL}_2(k)$, using the primitive idempotents which were constructed before by the author., Comment: 14 pages

In this paper we construct primitive idempotents of the hyperalgebra for the $r$-th Frobenius kernel of the algebraic group ${\rm SL}(2,k)$., Comment: 31 pages

Intra-operative 2D-3D registration of X-ray images with pre-operatively acquired CT scans is a crucial procedure in orthopedic surgeries. Anatomical landmarks pre-annotated in the CT volume can be detected in X-ray images to establish 2D-3D correspondences, which are then utilized for registration. However, registration often fails in certain view angles due to poor landmark visibility. We propose a novel method to address this issue by detecting arbitrary landmark points in X-ray images. Our approach represents 3D points as distinct subspaces, formed by feature vectors (referred to as ray embeddings) corresponding to intersecting rays. Establishing 2D-3D correspondences then becomes a task of finding ray embeddings that are close to a given subspace, essentially performing an intersection test. Unlike conventional methods for landmark estimation, our approach eliminates the need for manually annotating fixed landmarks. We trained our model using the synthetic images generated from CTPelvic1K CLINIC dataset, which contains 103 CT volumes, and evaluated it on the DeepFluoro dataset, comprising real X-ray images. Experimental results demonstrate the superiority of our method over conventional methods. The code is available at https://github.com/Pragyanstha/rayemb., Comment: Accepted as an oral presentation at ACCV 2024

We demonstrate that a quantum field theory (QFT) in general two-dimensional curved spacetimes can be realized by a system of quantum spins or qubits. We consider a spin-1/2 model on a one-dimensional ring with spatially and temporally varying exchange couplings and magnetic fields. This model reduces to a QFT of Majorana fermions in the continuum limit. From this correspondence, we establish a dictionary for translating between the spacetime-dependent parameters of the spin model and the general metric on which the QFT is defined. After addressing the general case, we consider the Friedmann-Lema\^{\i}tre-Robertson-Walker (FLRW) metric as a simple example. According to the dictionary, the QFT of Majorana fermions on the FLRW metric corresponds to the Ising model with a time-dependent transverse magnetic field. We demonstrate that the production of Majorana particles in the expanding universe can be simulated with the transverse-field Ising model by increasing the strength of the magnetic field. Furthermore, we examine the Unruh effect through the spin system by using our prescription and show the direct relation between the entanglement (or modular) Hamiltonian in the spin system and the Rindler Hamiltonian. This approach provides an experimentally viable system for probing various phenomena in QFT within curved spacetime, while also opening the door to uncovering nontrivial phenomena in spin systems inspired by curved spacetime physics. It offers fresh perspectives on both QFT in curved spacetimes and quantum many-body spin systems, revealing profound connections between these fields., Comment: 15 pages, 3 figures

The construction of highly coherent x-ray sources has enabled new research opportunities across the scientific landscape. The maximum raw data rate per beamline now exceeds 40 GB/s, posing unprecedented challenges for the online processing and offline storage of the big data. Such challenge is particularly prominent for x-ray photon correlation spectroscopy (XPCS), where real time analyses require simultaneous calculation on all the previously acquired data in the time series. We present a homomorphic compression scheme to effectively reduce the computational time and memory space required for XPCS analysis. Leveraging similarities in the mathematical expression between a matrix-based compression algorithm and the correlation calculation, our approach allows direct operation on the compressed data without their decompression. The lossy compression reduces the computational time by a factor of 10,000, enabling real time calculation of the correlation functions at kHz framerate. Our demonstration of a homomorphic compression of scientific data provides an effective solution to the big data challenge at coherent light sources. Beyond the example shown in this work, the framework can be extended to facilitate real-time operations directly on a compressed data stream for other techniques.

The Leggett-Garg inequality (LGI) serves as a criterion to determine the adherence of macroscopic system dynamics to macrorealism, as postulated by Leggett and Garg. A violation of this inequality implies either the absence of a realistic description of the system or the impracticality of noninvasive measurements. In this work, we investigate the violation of the LGI for the system of bosons in a double-well potential. Specifically, we explore the violation of the LGI in the dynamics of bosons in a double-well potential in the Bose-Einstein-Condensation (BEC) regime, where the system can be considered as two weakly coupled Bose condensates, and in the single-particle regime to establish the conditions under which the violation of the LGI occurs. Our analysis reveals that the LGI is violated due to Josephson oscillations, while it remains unviolated in the strong coupling regime, attributed to the self-trapping phenomena. Notably, we observe that the violation of the LGI becomes increasingly significant as the particle number increases. These findings provide valuable insights into the macrorealistic behavior of Bose condensates and highlight the effect of measurements on the dynamics of a macroscopic system., Comment: 14 pages, 10 figures

We introduce a novel laser-scanning optical microscopy technique that employs optical-frequency-comb (OFC) lasers. This method facilitates multimodal spectroscopic imaging by analyzing interferograms produced via a dual-comb spectroscopic approach. Such interferograms capture comprehensive light information, including amplitude, phase, polarization, frequency, and time of flight information, enabling multimodal imaging from a single measurement. We demonstrate the potential of this technique across several spectroscopic imaging applications., Comment: 11 pages, 8 figures

We numerically examine binary-sized granular mixtures confined between two parallel walls subjected to vertical vibration using the discrete element method. For a size ratio of $3$ between large and small particles, we study the structure of large particles in moderately dense regimes where the combined two-dimensional packing fractions of both particle sizes exceed $1$. When the fraction of small particles is small, segregation of the large particles occurs. In contrast, as the fraction of small particles increases, an effective repulsion between the large particles emerges over distances greater than the large particle diameter, suppressing their segregation. The emergence of reduction in segregation is confirmed for another size ratio, vibrational acceleration, system size, and for a case of bidisperse size distribution. Additionally, at the size ratio of $3$, the effective repulsion induces a hexagonal phase of the large particles at packing fractions lower than in mono-component systems. This work will provide a fresh insight into granular physics, prompting further experimental and theoretical study., Comment: 28 pages, 20 figures

Error-bounded lossy compression has been effective in significantly reducing the data storage/transfer burden while preserving the reconstructed data fidelity very well. Many error-bounded lossy compressors have been developed for a wide range of parallel and distributed use cases for years. These lossy compressors are designed with distinct compression models and design principles, such that each of them features particular pros and cons. In this paper we provide a comprehensive survey of emerging error-bounded lossy compression techniques for different use cases each involving big data to process. The key contribution is fourfold. (1) We summarize an insightful taxonomy of lossy compression into 6 classic compression models. (2) We provide a comprehensive survey of 10+ commonly used compression components/modules used in error-bounded lossy compressors. (3) We provide a comprehensive survey of 10+ state-of-the-art error-bounded lossy compressors as well as how they combine the various compression modules in their designs. (4) We provide a comprehensive survey of the lossy compression for 10+ modern scientific applications and use-cases. We believe this survey is useful to multiple communities including scientific applications, high-performance computing, lossy compression, and big data., Comment: submitted to ACM Computing journal, requited to be 35 pages including references

We numerically study the flow of dense granular materials between parallel plates driven by an external force. The granular materials form a jammed solid-like state when the external force is below a critical force, while they flow like fluids above the critical force. The transition is characterized by the mass flux. The critical force depends on the average packing fraction and the distance between the plates. The scaling laws for the critical force and the mass flux are predicted theoretically based on a continuum model. They are numerically verified.

We investigate spatially inhomogeneous solutions in a top-down holographic model: the D3/D7 model which provides a holographic description of the chiral phase transition for a finite external magnetic field, chemical potential, and temperature. We numerically find a domain wall (or kink) solution in the three dimensional space, which incorporates between the chiral symmetry broken phase at the spatial infinity, under the homogeneous sources. Along with the inhomogeneity of the chiral condensate, the charge density is also spatially modulated. The modulated charge density and finite magnetic field lead to the chiral edge current close to the domain wall. We explore the dependences of those profiles on the chemical potential and temperature near the first and second order phase transition points. Our results indicate that the inhomogeneous solutions we found are in good agreement with those obtained by the Ginzburg--Landau theory in the vicinity of the transition points., Comment: 30 pages, 20 figures; v2: version published in JHEP

An aqueous sodium chloride solution was injected at a controlled rate into a granular bed in a quasi-two-dimensional cell. The granular bed was made of dried, highly water-absorbent gel particles whose swelling rate was controlled by the salinity of the injected fluid. At a high salinity level (low swelling rate), high injection rate, and short timescale, the injected fluid percolated between the gel particles in an isotropic manner. Meanwhile, at a low salinity level (high swelling rate), low injection rate, and long timescale, the gel particles clogged the flow path, resulting in anisotropic branch-like structures of the injected fluid front. The transition of the injection pattern could be understood based on the ratio of the characteristic timescales of swelling and injection. Moreover, the clogged pattern showed an oscillatory pressure drop whose amplitude was increased with higher salinity. Such an oscillatory behavior observed in an injection process in a swelling gel particle may be relevant in geological situation; i.e., such as fluid migration underground., Comment: 6 pages, 4 figures

We construct an efficient momentum space approach to the superconductivity in quasiperiodic multilayer systems. To this end, we extend the Brillouin zone (BZ) folding method to the superconducting (SC) phases by formulating the gap equation in the momentum space representation with the BZ folding. In this picture, we find that the SC order parameters generally exhibit finite momenta, which corresponds to the spatial fluctuation of the SC order in quasiperiodic multilayers. We apply the BZ folding method to a one dimensional model of quasiperiodic superconductors and demonstrate its numerical efficiency compared to the conventional method based on the real space representation with a large system size., Comment: 11 pages, 8 figures

We give generating sets of the Jacobson radical of the hyperalgebra of the $r$-th Frobenius kernel of the algebraic group ${\rm SL}_2$ over an algebraically closed field of characteristic $p>0$. This result generalizes earlier work by Wong for $r=1$ and odd $p$., Comment: 35 pages. To appear in Journal of Computational Algebra

In the hyperalgebra of the $r$-th Frobenius kernel of a universal Chevalley group over a field of characteristic $p>0$, we study some subsets and the subalgebras generated by them and give some results. We are particularly interested in the case that $p$ is 'very small' and the Dynkin diagram is not simply-laced., Comment: 21 pages

The recent trend toward deep learning has led to the development of a variety of highly innovative AI accelerator architectures. One such architecture, the Cerebras Wafer-Scale Engine 2 (WSE-2), features 40 GB of on-chip SRAM, making it a potentially attractive platform for latency- or bandwidth-bound HPC simulation workloads. In this study, we examine the feasibility of performing continuous energy Monte Carlo (MC) particle transport on the WSE-2 by porting a key kernel from the MC transport algorithm to Cerebras's CSL programming model. New algorithms for minimizing communication costs and for handling load balancing are developed and tested. The WSE-2 is found to run 130 times faster than a highly optimized CUDA version of the kernel run on an NVIDIA A100 GPU -- significantly outpacing the expected performance increase given the difference in transistor counts between the architectures.

In the hyperalgebra $\mathcal{U}_r$ of the $r$-th Frobenius kernel $({\rm SL}_2)_r$ of the algebraic group ${\rm SL}_2$, we construct a basis of the $\mathcal{U}_r$-module generated by a certain element which was given by the author before. As its applications, we also prove some results on the $\mathcal{U}_r$-modules and the algebra $\mathcal{U}_r$., Comment: 30 pages

Let $G$ be a simply connected and simple algebraic group defined and split over a finite prime field $\mathbb{F}_p$ of $p$ elements. In this paper, using an $\mathbb{F}_p$-linear map splitting Frobenius endomorphism on a hyperalgebra relative to $G$, we obtain some $\mathbb{F}_p$-linear isomorphisms induced by multiplication in the hyperalgebra., Comment: 28 pages