arXiv subjects

Qiao Zhang

Publications and source records attributed to Qiao Zhang.

32 records · Page 2Linked to original sources

arXiv2024

Comet: A Communication-efficient and Performant Approximation for Private Transformer Inference

The prevalent use of Transformer-like models, exemplified by ChatGPT in modern language processing applications, underscores the critical need for enabling private inference essential for many cloud-based services reliant on such models. However, current privacy-preserving frameworks impose significant communication burden, especially for non-linear computation in Transformer model. In this paper, we introduce a novel plug-in method Comet to effectively reduce the communication cost without compromising the inference performance. We second introduce an efficient approximation method to eliminate the heavy communication in finding good initial approximation. We evaluate our Comet on Bert and RoBERTa models with GLUE benchmark datasets, showing up to 3.9$\times$ less communication and 3.5$\times$ speedups while keep competitive model performance compared to the prior art.

cs.LG↗

arXiv2024

Interpretable Models for Detecting and Monitoring Elevated Intracranial Pressure

Detecting elevated intracranial pressure (ICP) is crucial in diagnosing and managing various neurological conditions. These fluctuations in pressure are transmitted to the optic nerve sheath (ONS), resulting in changes to its diameter, which can then be detected using ultrasound imaging devices. However, interpreting sonographic images of the ONS can be challenging. In this work, we propose two systems that actively monitor the ONS diameter throughout an ultrasound video and make a final prediction as to whether ICP is elevated. To construct our systems, we leverage subject matter expert (SME) guidance, structuring our processing pipeline according to their collection procedure, while also prioritizing interpretability and computational efficiency. We conduct a number of experiments, demonstrating that our proposed systems are able to outperform various baselines. One of our SMEs then manually validates our top system's performance, lending further credibility to our approach while demonstrating its potential utility in a clinical setting.

eess.IV↗

arXiv2023

The Furstenberg property in Puiseux monoids

Let $M$ be a commutative monoid. The monoid $M$ is called atomic if every non-invertible element of $M$ factors into atoms (i.e., irreducible elements), while $M$ is called a Furstenberg monoid if every non-invertible element of $M$ is divisible by an atom. Additive submonoids of $\mathbb{Q}$ consisting of nonnegative rationals are called Puiseux monoids, and their atomic structure has been actively studied during the past few years. The primary purpose of this paper is to investigate the property of being Furstenberg in the context of Puiseux monoids. In this direction, we consider some properties weaker than being Furstenberg, and then we connect these properties with some atomic results which have been already established for Puiseux monoids.

math.AC↗

arXiv2023

The sum of root-leaf distance interdiction problem with cardinality constraint by upgrading edges on trees

A network for the transportation of supplies can be described as a rooted tree with a weight of a degree of congestion for each edge. We take the sum of root-leaf distance (SRD) on a rooted tree as the whole degree of congestion of the tree. Hence, we consider the SRD interdiction problem on trees with cardinality constraint by upgrading edges (denoted by (SDIPTC) in brief). It aims to maximize the SRD by upgrading the weights of $N$ critical edges such that the total upgrade cost under some measurement is upper-bounded by a given value. The relevant minimum cost problem (MCSDIPTC) aims to minimize the total upgrade cost on the premise that the SRD is lower-bounded by a given value. We develop two different norms including weighted $l_\infty$ norm and weighted bottleneck Hamming distance to measure the upgrade cost. We propose two binary search algorithms within O($n\log n$) time for the problems (SDIPTC) under the two norms, respectively. For problems (MCSDIPTC),we propose two binary search algorithms within O($N n^2$) and O($n \log n$) under weighted $l_\infty$ norm and weighted bottleneck Hamming distance, respectively. These problems are solved through their subproblems (SDIPT) and (MCSDIPT), in which we ignore the cardinality constraint on the number of upgraded edges. Finally, we design numerical experiments to show the effectiveness of these algorithms.

math.OC↗

arXiv2022

MobilePTX: Sparse Coding for Pneumothorax Detection Given Limited Training Examples

Point-of-Care Ultrasound (POCUS) refers to clinician-performed and interpreted ultrasonography at the patient's bedside. Interpreting these images requires a high level of expertise, which may not be available during emergencies. In this paper, we support POCUS by developing classifiers that can aid medical professionals by diagnosing whether or not a patient has pneumothorax. We decomposed the task into multiple steps, using YOLOv4 to extract relevant regions of the video and a 3D sparse coding model to represent video features. Given the difficulty in acquiring positive training videos, we trained a small-data classifier with a maximum of 15 positive and 32 negative examples. To counteract this limitation, we leveraged subject matter expert (SME) knowledge to limit the hypothesis space, thus reducing the cost of data collection. We present results using two lung ultrasound datasets and demonstrate that our model is capable of achieving performance on par with SMEs in pneumothorax identification. We then developed an iOS application that runs our full system in less than 4 seconds on an iPad Pro, and less than 8 seconds on an iPhone 13 Pro, labeling key regions in the lung sonogram to provide interpretable diagnoses.

cs.CV↗

arXiv2022

Lower Bounds for Rankin-Selberg $L$-functions on the Edge of the Critical Strip

Let $F$ be a number field, and let $π_1$ and $π_2$ be distinct unitary cuspidal automorphic representations of $\operatorname{GL}_{n_1}(\mathbb{A}_F)$ and $\operatorname{GL}_{n_2}(\mathbb{A}_F)$ respectively. In this paper, we derive new lower bounds for the Rankin-Selberg $L$-function $L(s, π_1 \times \widetildeπ_2)$ along the edge $\Re s = 1$ of the critical strip in the $t$-aspect. The corresponding zero-free region for $L(s, π_1 \times \widetildeπ_2)$ is also determined.

math.NT↗

arXiv2022

Joint Linear and Nonlinear Computation across Functions for Efficient Privacy-Preserving Neural Network Inference

While it is encouraging to witness the recent development in privacy-preserving Machine Learning as a Service (MLaaS), there still exists a significant performance gap for its deployment in real-world applications. We observe the state-of-the-art frameworks follow a compute-and-share principle for every function output where the summing in linear functions, which is the last of two steps for function output, involves all rotations (which is the most expensive HE operation), and the multiplexing in nonlinear functions, which is also the last of two steps for function output, introduces noticeable communication rounds. Therefore, we challenge the conventional compute-and-share logic and introduce the first joint linear and nonlinear computation across functions that features by 1) the PHE triplet for computing the nonlinear function, with which the multiplexing is eliminated; 2) the matrix encoding to calculate the linear function, with which all rotations for summing is removed; and 3) the network adaptation to reassemble the model structure, with which the joint computation module is utilized as much as possible. The boosted efficiency is verified by the numerical complexity, and the experiments demonstrate up to 13x speedup for various functions used in the state-of-the-art models and up to 5x speedup over mainstream neural networks.

cs.CR↗

arXiv2022

High-responsivity MoS$_2$ hot-electron telecom-band photodetector integrated with microring resonator

We report a high-responsive hot-electron photodetector based on the integration of an Au-MoS$_2$ junction with a silicon nitride microring resonator (MRR) for detecting telecom-band light. The coupling of the evanescent field of the silicon nitride MRR with the Au-MoS$_2$ Schottky junction region enhances the hot-electron injection efficiency. The device exhibits a high responsivity of 154.6 mA W-1 at the wavelength of 1516 nm, and the moderately uniform responsivities are obtained over the wavelength range of 1500 nm-1630 nm. This MRR-enhanced MoS2 hot-electron photodetector offers possibilities for integrated optoelectronic systems.

physics.optics↗

arXiv2022

Do it Like the Doctor: How We Can Design a Model That Uses Domain Knowledge to Diagnose Pneumothorax

Computer-aided diagnosis for medical imaging is a well-studied field that aims to provide real-time decision support systems for physicians. These systems attempt to detect and diagnose a plethora of medical conditions across a variety of image diagnostic technologies including ultrasound, x-ray, MRI, and CT. When designing AI models for these systems, we are often limited by little training data, and for rare medical conditions, positive examples are difficult to obtain. These issues often cause models to perform poorly, so we needed a way to design an AI model in light of these limitations. Thus, our approach was to incorporate expert domain knowledge into the design of an AI model. We conducted two qualitative think-aloud studies with doctors trained in the interpretation of lung ultrasound diagnosis to extract relevant domain knowledge for the condition Pneumothorax. We extracted knowledge of key features and procedures used to make a diagnosis. With this knowledge, we employed knowledge engineering concepts to make recommendations for an AI model design to automatically diagnose Pneumothorax.

cs.AI↗

arXiv2021

GALA: Greedy ComputAtion for Linear Algebra in Privacy-Preserved Neural Networks

Machine Learning as a Service (MLaaS) is enabling a wide range of smart applications on end devices. However, privacy-preserved computation is still expensive. Our investigation has found that the most time-consuming component of the HE-based linear computation is a series of Permutation (Perm) operations that are imperative for dot product and convolution in privacy-preserved MLaaS. To this end, we propose GALA: Greedy computAtion for Linear Algebra in privacy-preserved neural networks, which views the HE-based linear computation as a series of Homomorphic Add, Mult and Perm operations and chooses the least expensive operation in each linear computation step to reduce the overall cost. GALA makes the following contributions: (1) It introduces a row-wise weight matrix encoding and combines the share generation that is needed for the GC-based nonlinear computation, to reduce the Perm operations for the dot product; (2) It designs a first-Add-second-Perm approach (named kernel grouping) to reduce Perm operations for convolution. As such, GALA efficiently reduces the cost for the HE-based linear computation, which is a critical building block in almost all of the recent frameworks for privacy-preserved neural networks, including GAZELLE (Usenix Security'18), DELPHI (Usenix Security'20), and CrypTFlow2 (CCS'20). With its deep optimization of the HE-based linear computation, GALA can be a plug-and-play module integrated into these systems to further boost their efficiency. Our experiments show that it achieves a significant speedup up to 700x for the dot product and 14x for the convolution computation under different data dimensions. Meanwhile, GALA demonstrates an encouraging runtime boost by 2.5x, 2.7x, 3.2x, 8.3x, 7.7x, and 7.5x over GAZELLE and 6.5x, 6x, 5.7x, 4.5x, 4.2x, and 4.1x over CrypTFlow2, on AlexNet, VGG, ResNet-18, ResNet-50, ResNet-101, and ResNet-152, respectively.

cs.CR↗

arXiv2021

CHEETAH: An Ultra-Fast, Approximation-Free, and Privacy-Preserved Neural Network Framework based on Joint Obscure Linear and Nonlinear Computations

Machine Learning as a Service (MLaaS) is enabling a wide range of smart applications on end devices. However, such convenience comes with a cost of privacy because users have to upload their private data to the cloud. This research aims to provide effective and efficient MLaaS such that the cloud server learns nothing about user data and the users cannot infer the proprietary model parameters owned by the server. This work makes the following contributions. First, it unveils the fundamental performance bottleneck of existing schemes due to the heavy permutations in computing linear transformation and the use of communication intensive Garbled Circuits for nonlinear transformation. Second, it introduces an ultra-fast secure MLaaS framework, CHEETAH, which features a carefully crafted secret sharing scheme that runs significantly faster than existing schemes without accuracy loss. Third, CHEETAH is evaluated on the benchmark of well-known, practical deep networks such as AlexNet and VGG-16 on the MNIST and ImageNet datasets. The results demonstrate more than 100x speedup over the fastest GAZELLE (Usenix Security'18), 2000x speedup over MiniONN (ACM CCS'17) and five orders of magnitude speedup over CryptoNets (ICML'16). This significant speedup enables a wide range of practical applications based on privacy-preserved deep neural networks.

cs.LG↗

arXiv2019

Novel Balanced Single/Dual-band Bandpass Filters Based on Circular Patch Resonator

A new approach for designing the single-band and dual-band balanced bandpass filters (BPFs) based on the circular patch resonator is proposed in this work. By modifying the output balanced ports of the balanced BPFs to inhibit or transport TM31 mode of the circular patch resonator under the differential-mode (DM) excitation, a single-band or dual-band balanced BPF can realized. As the odd modes could be restrained under the common-mode (CM) excitation, the balanced BPFs could achieve high CM suppression in the DM passbands. Meanwhile, several perturbing ways are introduced to improve the performance of the balanced BPFs. Specifically, perturbation vias are introduced to disturb the electric field distribution to improve the CM suppression in the DM passbands, and slots are cut on the circular patch resonators to perturb the current distribution to modify the second DM passband of the dual-band balanced BPF. Furthermore, the proposed single-band and dual-band balanced BPFs are fabricated and characterized. Good agreement can be achieved between the simulated and measured results of the fabricated balanced BPFs, which validates the feasibility of the proposed design method.

eess.SY↗

arXiv2018

Volur: Concurrent Edge/Core Route Control in Data Center Networks

A perennial question in computer networks is where to place functionality among components of a distributed computer system. In data centers, one option is to move all intelligence to the edge, essentially relegating switches and middleboxes, regardless of their programmability, to simple static routing policies. Another is to add more intelligence to the middle of the network in the hopes that it can handle any issue that arises. This paper presents an architecture, called Volur, that provides a third option by facilitating the co-existence of an intelligent network with an intelligent edge. The key architectural principle of Volur is predictability of the network. We describe the key design requirements, and show through case studies how our approach facilitates more democratic innovation of all parts of the network. We also demonstrate the practicality of our architecture by describing how to implement the architecture on top of existing hardware and by deploying a prototype on top of a large production data center.

cs.NI↗

arXiv2014

Fourier Coefficients of Theta Functions at Cusps other than Infinity

In this paper we study the Fourier coefficients of theta functions attached to Dirichlet characters at cusps other than infinity. The method is based on expressing them in terms of explicit elements of the adelic Schwartz space and studying the action of the adelic metaplectic group on these elements. We derive explicit formulae for the Fourier coefficients at all cusps. For the sake of simplicity, some restrictions are placed on the Dirichlet characters considered.

math.NT↗