RSS 1.0Preprints
Recent Submissions
-
Optimal RIS Partitioning and Power Control for Bidirectional NOMA Networks(Institute of Electrical and Electronics Engineers (IEEE), 2023-03-27) [Preprint]This study delves into the capabilities of reconfig- urable intelligent surfaces (RISs) in enhancing bidirectional non- orthogonal multiple access (NOMA) networks. The proposed approach partitions RIS to optimize the channel conditions for NOMA users, improving NOMA gain and eliminating the re- quirement for uplink (UL) power control. The proposed approach is rigorously evaluated under four practical operational regimes; 1) Quality-of-Service (QoS) sufficient regime, 2) RIS and power efficient regime, 3) max-min fair regime, and 4) maximum throughput regime, each subject to both UL and downlink (DL) QoS constraints. By leveraging decoupled nature of RIS portions and base station (BS) transmit power, closed-form solutions are derived to demonstrate how optimal RIS partitioning can meet UL-QoS requirements while optimal BS power control can ensure DL-QoS compliance. Our analytical findings are validated through simulations, highlighting the significant benefits that RISs can bring to the NOMA networks in the aforementioned operational scenarios.
-
Integrated Nested Laplace Approximations for Large-Scale Spatial-Temporal Bayesian Modeling(arXiv, 2023-03-27) [Preprint]Bayesian inference tasks continue to pose a computational challenge. This especially holds for spatial-temporal modeling where high-dimensional latent parameter spaces are ubiquitous. The methodology of integrated nested Laplace approximations (INLA) provides a framework for performing Bayesian inference applicable to a large subclass of additive Bayesian hierarchical models. In combination with the stochastic partial differential equations (SPDE) approach it gives rise to an efficient method for spatial-temporal modeling. In this work we build on the INLA-SPDE approach, by putting forward a performant distributed memory variant, INLA-DIST, for large-scale applications. To perform the arising computational kernel operations, consisting of Cholesky factorizations, solving linear systems, and selected matrix inversions, we present two numerical solver options, a sparse CPU-based library and a novel blocked GPU-accelerated approach which we propose. We leverage the recurring nonzero block structure in the arising precision (inverse covariance) matrices, which allows us to employ dense subroutines within a sparse setting. Both versions of INLA-DIST are highly scalable, capable of performing inference on models with millions of latent parameters. We demonstrate their accuracy and performance on synthetic as well as real-world climate dataset applications.
-
Towards black-box parameter estimation(arXiv, 2023-03-27) [Preprint]Deep learning algorithms have recently shown to be a successful tool in estimating parameters of statistical models for which simulation is easy, but likelihood computation is challenging. But the success of these approaches depends on simulating parameters that sufficiently reproduce the observed data, and, at present, there is a lack of efficient methods to produce these simulations. We develop new black-box procedures to estimate parameters of statistical models based only on weak parameter structure assumptions. For well-structured likelihoods with frequent occurrences, such as in time series, this is achieved by pre-training a deep neural network on an extensive simulated database that covers a wide range of data sizes. For other types of complex dependencies, an iterative algorithm guides simulations to the correct parameter region in multiple rounds. These approaches can successfully estimate and quantify the uncertainty of parameters from non-Gaussian models with complex spatial and temporal dependencies. The success of our methods is a first step towards a fully flexible automatic black-box estimation framework.
-
A Stable Neural Network-Based Eikonal Tomography using Hard-Constrained Measurements(Authorea, Inc., 2023-03-26) [Preprint]Eikonal tomography, or travel time inversion, has been one of the primary seismological tools for decades and has been used to understand Earth’s properties and dynamic processes. At the heart of the inversion process is the need for an accurate, and preferably flexible, eikonal solver to compute the travel time field. Most of the conventional eikonal solvers, however, suffer from first-order convergence errors and difficulties in dealing with irregular computational grids. Physics-informed neural networks (PINNs) have been introduced to tackle these problems and have shown great success in addressing those challenges. Nevertheless, these approaches still suffer from slow convergence and unstable training dynamics due to the multi-term nature of the loss function. To improve on this, we propose a new formulation for the isotropic eikonal equation, which imposes boundary conditions as hard constraints. We employ the theory of functional connections to the eikonal tomography problem, which allows for the utilization of a single loss term for training the PINN model. Through rigorous numerical tests, its efficiency, stability, and flexibility in tackling a variety of cases, including topography-dependent and 3D models, are attested, thus providing an efficient and stable PINN-based eikonal tomography.
-
BlobGAN-3D: A Spatially-Disentangled 3D-Aware Generative Model for Indoor Scenes(arXiv, 2023-03-26) [Preprint]3D-aware image synthesis has attracted increasing interest as it models the 3D nature of our real world. However, performing realistic object-level editing of the generated images in the multi-object scenario still remains a challenge. Recently, a 2D GAN termed BlobGAN has demonstrated great multi-object editing capabilities on real-world indoor scene datasets. In this work, we propose BlobGAN-3D, which is a 3D-aware improvement of the original 2D BlobGAN. We enable explicit camera pose control while maintaining the disentanglement for individual objects in the scene by extending the 2D blobs into 3D blobs. We keep the object-level editing capabilities of BlobGAN and in addition allow flexible control over the 3D location of the objects in the scene. We test our method on real-world indoor datasets and show that our method can achieve comparable image quality compared to the 2D BlobGAN and other 3D-aware GAN baselines while being able to enable camera pose control and object-level editing in the challenging multi-object real-world scenarios.
-
On the effect of different samplings to the solution of parametric PDE Eigenvalue Problems(arXiv, 2023-03-25) [Preprint]In this article we apply reduced order techniques for the approximation of parametric eigenvalue problems. The effect of the choice of sampling points is investigated. Here we use the standard proper orthogonal decomposition technique to obtain the basis of the reduced space and Galerking orthogonal technique is used to get the reduced problem. We present some numerical results and observe that the use of sparse sampling is a good idea for sampling if the dimension of parameter space is high.
-
A Scalable Laser-Based Underwater Wireless Optical Communication Solution between Autonomous Underwater Vehicle Fleets(Elsevier BV, 2023-03-23) [Preprint]The development of multiple autonomous underwater vehicles (AUVs) has changed the way of relying on a single and expensive AUV to conduct underwater surveys, which is becoming increasingly attractive to marine researchers. Communication between AUV fleets is an urgent concern due to the data rate limitation of underwater acoustic communication. Laser-based underwater wireless optical communication (UWOC) is a potential solution once the link-establishing requirement between AUVs can be met. Due to the limited coverage area of the laser beam, the previous pointing, acquisition, and tracking (PAT) method is to quickly adjust the beam direction and search for the target according to the set scanning path. We propose a scalable laser-based link establishment method that combines the maneuvering of the AUV, the acoustic positioning, and the control of the optical system for increased efficiency. Our approach outperformed the existing PAT approach in a simulation environment and successfully established laser links. In actual machine experiments, the results proved that the proposed approach can be implemented in practical scenarios.
-
Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs(arXiv, 2023-03-23) [Preprint]In this paper we investigate the frequency sensitivity of Deep Neural Networks (DNNs) when presented with clean samples versus poisoned samples. Our analysis shows significant disparities in frequency sensitivity between these two types of samples. Building on these findings, we propose FREAK, a frequency-based poisoned sample detection algorithm that is simple yet effective. Our experimental results demonstrate the efficacy of FREAK not only against frequency backdoor attacks but also against some spatial attacks. Our work is just the first step in leveraging these insights. We believe that our analysis and proposed defense mechanism will provide a foundation for future research and development of backdoor defenses.
-
5G-Aided RTK Positioning in GNSS-Deprived Environments(arXiv, 2023-03-23) [Preprint]This paper considers the localization problem in a 5G-aided global navigation satellite system (GNSS) based on real-time kinematic (RTK) technique. Specifically, the user's position is estimated based on the hybrid measurements, including GNSS pseudo-ranges, GNSS carrier phases, 5G angle-of-departures, and 5G channel delays. The underlying estimation problem is solved by steps that comprise obtaining the float solution, ambiguity resolution, and resolving the fixed solution. The analysis results show that the involvement of 5G observations can enable localization under satellite-deprived environments, inclusive of extreme cases with only 2 or 3 visible satellites. Moreover, extensive simulation results reveal that with the help of 5G observations, the proposed algorithm can significantly reduce the estimation error of the user's position and increase the success rate of carrier-phase ambiguity resolution.
-
Spatial Path Index Modulation in mmWave/THz-Band Integrated Sensing and Communications(arXiv, 2023-03-22) [Preprint]As the demand for wireless connectivity continues to soar, the fifth generation and beyond wireless networks are exploring new ways to efficiently utilize the wireless spectrum and reduce hardware costs. One such approach is the integration of sensing and communications (ISAC) paradigms to jointly access the spectrum. Recent ISAC studies have focused on upper millimeter-wave and low terahertz bands to exploit ultrawide bandwidths. At these frequencies, hybrid beamformers that employ fewer radio-frequency chains are employed to offset expensive hardware but at the cost of lower multiplexing gains. Wideband hybrid beamforming also suffers from the beam-split effect arising from the subcarrier-independent (SI) analog beamformers. To overcome these limitations, this paper introduces a spatial path index modulation (SPIM) ISAC architecture, which transmits additional information bits via modulating the spatial paths between the base station and communications users. We design the SPIM-ISAC beamformers by first estimating both radar and communications parameters by developing beam-split-aware algorithms. Then, we propose to employ a family of hybrid beamforming techniques such as hybrid, SI, and subcarrier-dependent analog-only, and beam-split-aware beamformers. Numerical experiments demonstrate that the proposed SPIM-ISAC approach exhibits significantly improved spectral efficiency performance in the presence of beam-split than that of even fully digital non-SPIM beamformers.
-
AB-Gen: Antibody Library Design with Generative Pre-trained Transformer and Deep Reinforcement Learning(Cold Spring Harbor Laboratory, 2023-03-21) [Preprint]Antibody leads must fulfill multiple desirable properties to be clinical candidates. Primarily due to the low throughput in the experimental procedure, the need for such multi-property optimization causes the bottleneck in preclinical antibody discovery and development, because addressing one issue usually causes another. We developed a reinforcement learning (RL) method, named AB-Gen, for antibody library design using a generative pre-trained Transformer (GPT) as the policy network of the RL agent. We showed that this model can learn the antibody space of heavy chain complementarity determining region 3 (CDRH3) and generate sequences with similar property distributions. Besides, when using HER2 as the target, the agent model of AB-Gen was able to generate novel CDRH3 sequences that fulfill multi-property constraints. 509 generated sequences were able to pass all property filters and three highly conserved residues were identified. The importance of these residues was further demonstrated by molecular dynamics simulations, which consolidated that the agent model was capable of grasping important information in this complex optimization task. Overall, the AB-Gen method is able to design novel antibody sequences with an improved success rate than the traditional propose-then-filter approach. It has the potential to be used in practical antibody design, thus empowering the antibody discovery and development process.
-
SEEING THROUGH THE CO2 PLUME: JOINT INVERSION-SEGMENTATION OF THE SLEIPNER 4D SEISMIC DATASET(arXiv, 2023-03-21) [Preprint]4D seismic inversion is the leading method to quantitatively monitor fluid flow dynamics in the subsurface, with applications ranging from enhanced oil recovery to subsurface CO2 storage. The process of inverting seismic data for reservoir properties is, however, a notoriously ill-posed inverse problem due to the band-limited and noisy nature of seismic data. This comes with additional challenges for 4D applications, given inaccuracies in the repeatability of the time-lapse acquisition surveys. Consequently, adding prior information to the inversion process in the form of properly crafted regularization terms is essential to obtain geologically meaningful subsurface models. Motivated by recent advances in the field of convex optimization, we propose a joint inversion-segmentation algorithm for 4D seismic inversion, which integrates Total-Variation and segmentation priors as a way to counteract the missing frequencies and noise present in 4D seismic data. The proposed inversion framework is applied to a pair of surveys from the open Sleipner 4D Seismic Dataset. Our method presents three main advantages over state-of-the-art least-squares inversion methods: 1. it produces high-resolution baseline and monitor acoustic models, 2. by leveraging similarities between multiple data, it mitigates the non-repeatable noise and better highlights the real time-lapse changes, and 3. it provides a volumetric classification of the acoustic impedance 4D difference model (time-lapse changes) based on user-defined classes, i.e., percentages of speed-up or slow-down in the subsurface. Such advantages may enable more robust stratigraphic/structural and quantitative 4D seismic interpretation and provide more accurate inputs for dynamic reservoir simulations. Alongside presenting our novel inversion method, in this work, we introduce a streamlined data pre-processing sequence for the 4D Sleipner post-stack seismic dataset, which includes time-shift estimation and well-to-seismic tie. Finally, we provide insights into the open-source framework for large-scale optimization that we used to implement the proposed algorithm in an efficient and scalable manner.
-
Engineering grain boundaries in monolayer molybdenum disulfide for an efficient water/ion separation(Research Square Platform LLC, 2023-03-20) [Preprint]Atomically thin two-dimensional (2D) materials have long been considered as ideal platforms for developing separation membranes. However, it is difficult to generate uniform subnanometer pores over large areas on 2D materials. Herein, we report that the well-defined defect structure of monolayer MoS2, namely, eight-membered ring (8-MR) pores typically formed at the boundaries of two antiparallel grains, can serve as molecular sieves for efficient water/ion separation. The 8-MR pores (4.2 × 2.4 Å) in monolayer MoS2 allow rapid single-file water transport while rejecting various hydrated ions. Further, the density of grain boundaries and, consequently, the density of pores can be tuned by regulating the nucleation density and size of MoS2 grains during the chemical vapor deposition process. The optimized MoS2 membrane exhibited an ultrahigh water/NaCl selectivity of ~6.5 × 104 at a water permeance of 232 mol m−2 h−1 bar−1, outperforming the state-of-the-art desalination membranes. When used for direct hydrogen production from seawater by combining the forward osmosis and electrochemical water splitting processes, the membrane achieved ~40 times the energy conversion efficiency of commercial polymeric membranes. It also exhibited a rapid and selective proton transport behavior desirable for fuel cells and electrolysis. The bottom-up approach of creating precise pore structures on atomically thin films via grain boundary engineering presents a promising route for producing large-area membranes suitable for various applications.
-
Computationally Budgeted Continual Learning: What Does Matter?(arXiv, 2023-03-20) [Preprint]Continual Learning (CL) aims to sequentially train models on streams of incoming data that vary in distribution by preserving previous knowledge while adapting to new data. Current CL literature focuses on restricted access to previously seen data, while imposing no constraints on the computational budget for training. This is unreasonable for applications in-the-wild, where systems are primarily constrained by computational and time budgets, not storage. We revisit this problem with a large-scale benchmark and analyze the performance of traditional CL approaches in a compute-constrained setting, where effective memory samples used in training can be implicitly restricted as a consequence of limited computation. We conduct experiments evaluating various CL sampling strategies, distillation losses, and partial fine-tuning on two large-scale datasets, namely ImageNet2K and Continual Google Landmarks V2 in data incremental, class incremental, and time incremental settings. Through extensive experiments amounting to a total of over 1500 GPU-hours, we find that, under compute-constrained setting, traditional CL approaches, with no exception, fail to outperform a simple minimal baseline that samples uniformly from memory. Our conclusions are consistent in a different number of stream time steps, e.g., 20 to 200, and under several computational budgets. This suggests that most existing CL methods are particularly too computationally expensive for realistic budgeted deployment.
-
Finite element discretization of a biological network formation system: a preliminary study(arXiv, 2023-03-19) [Preprint]A finite element discretization is developed for the Cai-Hu model, describing the formation of biological networks. The model consists of a non linear elliptic equation for the pressure p and a non linear reaction-diffusion equation for the conductivity tensor C. The problem requires high resolution due to the presence of multiple scales, the stiffness in all its components and the non linearities. We propose a low order finite element discretization in space coupled with a semi-implicit time advancing scheme. The code is validated with several numerical tests performed with various choices for the parameters involved in the system. In absence of the exact solution, we apply Richardson extrapolation technique to estimate the order of the method.
-
Benchmark modeling and 3D applications of solidification and macro-segregation based on an operator-splitting and fully decoupled scheme with term-wise matrix assembly(arXiv, 2023-03-19) [Preprint]The solidification and macro-segregation problem involving unsteady multi-physics and multi-phase fields is typically a complex process with mass, momentum, heat, and species transfers among solid, mushy, and liquid phase regions. The quantitative prediction of phase change, chemical heterogeneities, and multi-phase and multi-component flows plays critical roles in many natural scenarios and industrial applications that involve many disciplines, like material, energy, and even planet science. In view of this, some scholars and research institutions have called for more contributors to join the benchmark analysis of solidification and segregation problems. Our work proposes an operator-splitting and matrix-based method to avoid non-linear systems. Also, the combination of vectorization and forward equation-based matrix assembly techniques enhances the implementability of extensions of 3D applications. Lastly, the novel scheme is well validated through a bunch of 2D and 3D benchmark cases. The numerical results also illustrate that this method can ensure accurate prediction and adequately capture the physical details of phenomena caused by the solutally and thermally driven flow, which include channel segregation, the formation of freckles, edge effect, aspect ratio effect, and 3D effect.
-
Generalization of the Orthodiagonal Involutive Type of Kokotsakis Flexible Polyhedra(arXiv, 2023-03-19) [Preprint]In this paper we introduce and study a remarkable class of mechanisms formed by a 3×3 arrangement of rigid and skew quadrilateral faces with revolute joints at the common edges. These Kokotsakis-type mechanisms with a quadrangular base and non-planar faces are a generalization of Izmestiev's orthodiagonal involutive type of Kokotsakis polyhedra formed by planar quadrilateral faces. Our algebraic approach yields a complete characterization of all complexes of the orthodiagonal involutive type. It is shown that one has 8 degrees of freedom to construct such mechanisms. This is illustrated by several examples, including cases that are not possible with planar faces.
-
LEP-AD: Language Embedding of Proteins and Attention to Drugs predicts drug target interactions(Cold Spring Harbor Laboratory, 2023-03-15) [Preprint]Predicting drug-target interactions is a tremendous challenge for drug development and lead optimization. Recent advances include training algorithms to learn drug-target interactions from data and molecular simulations. Here we utilize Evolutionary Scale Modeling (ESM-2) models to establish a Transformer protein language model for drug-target interaction predictions. Our architecture, LEP- AD, combines pre-trained ESM-2 and Transformer-GCN models predicting bind-ing affinity values. We report new best-in-class state-of-the-art results compared to competing methods such as SimBoost, DeepCPI, Attention-DTA, GraphDTA, and more using multiple datasets, including Davis, KIBA, DTC, Metz, ToxCast, and STITCH. Finally, we find that a pre-trained model with embedding of proteins (the LED-AD) outperforms a model using an explicit alpha-fold 3D representation of proteins (e.g., LEP-AD supervised by Alphafold). The LEP-AD model scales favorably in performance with the size of training data.
-
DualMix: Unleashing the Potential of Data Augmentation for Online Class-Incremental Learning(arXiv, 2023-03-14) [Preprint]Online Class-Incremental (OCI) learning has sparked new approaches to expand the previously trained model knowledge from sequentially arriving data streams with new classes. Unfortunately, OCI learning can suffer from catastrophic forgetting (CF) as the decision boundaries for old classes can become inaccurate when perturbated by new ones. Existing literature have applied the data augmentation (DA) to alleviate the model forgetting, while the role of DA in OCI has not been well understood so far. In this paper, we theoretically show that augmented samples with lower correlation to the original data are more effective in preventing forgetting. However, aggressive augmentation may also reduce the consistency between data and corresponding labels, which motivates us to exploit proper DA to boost the OCI performance and prevent the CF problem. We propose the Enhanced Mixup (EnMix) method that mixes the augmented samples and their labels simultaneously, which is shown to enhance the sample diversity while maintaining strong consistency with corresponding labels. Further, to solve the class imbalance problem, we design an Adaptive Mixup (AdpMix) method to calibrate the decision boundaries by mixing samples from both old and new classes and dynamically adjusting the label mixing ratio. Our approach is demonstrated to be effective on several benchmark datasets through extensive experiments, and it is shown to be compatible with other replay-based techniques.
-
Multigateway precoded NOMA in multibeam satellite multicast systems(Journal of Communications and Networks, Institute of Electrical and Electronics Engineers (IEEE), 2023-03-14) [Article]In this paper, we consider multigateway-based multibeam satellite non-orthogonal multiple access (NOMA) multicast systems. We investigate the improvement in spectral efficiency and accessibility by superimposing multiple signals on each beam at the same frequency and time resource employing NOMA, where multiple gateways transmit precoded signals to alleviate inter-beam interference caused by full frequency reuse. We formulate an optimization problem of maximizing the sum rate while satisfying the gateways and satellite transmission power constraints, where the precoding vector and power allocation for superposition coding as well as the decoding order for successive interference cancellation are optimized. This optimization problem is challenging to solve due to its non-convex mixed-integer nonlinear programming. However, a suboptimal solution can be obtained using a block coordinate descent algorithm. The simulation results of the proposed NOMA technique are compared with those of the orthogonal multiple access (OMA) technique. The proposed technique outperforms the OMA technique. We also investigate the impact of channel imperfection and decoding capability of the proposed algorithm through some selected simulation results.
