arXiv subjects

Yiming Chen

Publications and source records attributed to Yiming Chen.

At least 127 records · Page 7Linked to original sources

arXiv2024

TL-nvSRAM-CIM: Ultra-High-Density Three-Level ReRAM-Assisted Computing-in-nvSRAM with DC-Power Free Restore and Ternary MAC Operations

Accommodating all the weights on-chip for large-scale NNs remains a great challenge for SRAM based computing-in-memory (SRAM-CIM) with limited on-chip capacity. Previous non-volatile SRAM-CIM (nvSRAM-CIM) addresses this issue by integrating high-density single-level ReRAMs on the top of high-efficiency SRAM-CIM for weight storage to eliminate the off-chip memory access. However, previous SL-nvSRAM-CIM suffers from poor scalability for an increased number of SL-ReRAMs and limited computing efficiency. To overcome these challenges, this work proposes an ultra-high-density three-level ReRAMs-assisted computing-in-nonvolatile-SRAM (TL-nvSRAM-CIM) scheme for large NN models. The clustered n-selector-n-ReRAM (cluster-nSnRs) is employed for reliable weight-restore with eliminated DC power. Furthermore, a ternary SRAM-CIM mechanism with differential computing scheme is proposed for energy-efficient ternary MAC operations while preserving high NN accuracy. The proposed TL-nvSRAM-CIM achieves 7.8x higher storage density, compared with the state-of-art works. Moreover, TL-nvSRAM-CIM shows up to 2.9x and 1.9x enhanced energy-efficiency, respectively, compared to the baseline designs of SRAM-CIM and ReRAM-CIM, respectively.

cs.AR↗

arXiv2023

Progressive Poisoned Data Isolation for Training-time Backdoor Defense

Deep Neural Networks (DNN) are susceptible to backdoor attacks where malicious attackers manipulate the model's predictions via data poisoning. It is hence imperative to develop a strategy for training a clean model using a potentially poisoned dataset. Previous training-time defense mechanisms typically employ an one-time isolation process, often leading to suboptimal isolation outcomes. In this study, we present a novel and efficacious defense method, termed Progressive Isolation of Poisoned Data (PIPD), that progressively isolates poisoned data to enhance the isolation accuracy and mitigate the risk of benign samples being misclassified as poisoned ones. Once the poisoned portion of the dataset has been identified, we introduce a selective training process to train a clean model. Through the implementation of these techniques, we ensure that the trained model manifests a significantly diminished attack success rate against the poisoned data. Extensive experiments on multiple benchmark datasets and DNN models, assessed against nine state-of-the-art backdoor attacks, demonstrate the superior performance of our PIPD method for backdoor defense. For instance, our PIPD achieves an average True Positive Rate (TPR) of 99.95% and an average False Positive Rate (FPR) of 0.06% for diverse attacks over CIFAR-10 dataset, markedly surpassing the performance of state-of-the-art methods.

cs.CR↗

arXiv2023

MVControl: Adding Conditional Control to Multi-view Diffusion for Controllable Text-to-3D Generation

We introduce MVControl, a novel neural network architecture that enhances existing pre-trained multi-view 2D diffusion models by incorporating additional input conditions, e.g. edge maps. Our approach enables the generation of controllable multi-view images and view-consistent 3D content. To achieve controllable multi-view image generation, we leverage MVDream as our base model, and train a new neural network module as additional plugin for end-to-end task-specific condition learning. To precisely control the shapes and views of generated images, we innovatively propose a new conditioning mechanism that predicts an embedding encapsulating the input spatial and view conditions, which is then injected to the network globally. Once MVControl is trained, score-distillation (SDS) loss based optimization can be performed to generate 3D content, in which process we propose to use a hybrid diffusion prior. The hybrid prior relies on a pre-trained Stable-Diffusion network and our trained MVControl for additional guidance. Extensive experiments demonstrate that our method achieves robust generalization and enables the controllable generation of high-quality 3D content. Code available at https://github.com/WU-CVGL/MVControl/.

cs.CV↗

arXiv2023

ET3D: Efficient Text-to-3D Generation via Multi-View Distillation

Recent breakthroughs in text-to-image generation has shown encouraging results via large generative models. Due to the scarcity of 3D assets, it is hardly to transfer the success of text-to-image generation to that of text-to-3D generation. Existing text-to-3D generation methods usually adopt the paradigm of DreamFusion, which conducts per-asset optimization by distilling a pretrained text-to-image diffusion model. The generation speed usually ranges from several minutes to tens of minutes per 3D asset, which degrades the user experience and also imposes a burden to the service providers due to the high computational budget. In this work, we present an efficient text-to-3D generation method, which requires only around 8 $ms$ to generate a 3D asset given the text prompt on a consumer graphic card. The main insight is that we exploit the images generated by a large pre-trained text-to-image diffusion model, to supervise the training of a text conditioned 3D generative adversarial network. Once the network is trained, we are able to efficiently generate a 3D asset via a single forward pass. Our method requires no 3D training data and provides an alternative approach for efficient text-to-3D generation by distilling pre-trained image diffusion models.

cs.CV↗

arXiv2023

Comments on the double cone wormhole

In this paper we revisit the double cone wormhole introduced by Saad, Shenker and Stanford (SSS), which was shown to reproduce the ramp in the spectral form factor. As a first approximation we can say that this solution computes $\textrm{Tr}[e^{-iKT}]$, a trace of the "evolution" operator that generates Schwarzschild time translations on the two sided wormhole geometry. This point of view leads to a simple way to compute the normalization factor of the wormhole. When we have bulk matter fields, SSS suggested using a modified evolution $\tilde K$ which involves a slightly complex geometry, so that we are really computing $\textrm{Tr}[e^{-i\tilde{K}T}]$. We argue that, for general black holes, the spectrum of $\tilde K$ is given by quasinormal mode frequencies. We explain that this reproduces various features that were previously predicted from the spectral form factor on hydrodynamics grounds. We also give a general algebraic construction of the modified boost in terms of operators constructed from half sided modular inclusions. For the special case of JT gravity, we work out the backreaction of matter on the geometry of the double cone and find that it deforms the geometry in an undesirable direction. We finally give some comments on the possible physical interpretation of $\tilde K$.

hep-th↗

arXiv2023

Enhancing Black-Box Few-Shot Text Classification with Prompt-Based Data Augmentation

Training or finetuning large-scale language models (LLMs) such as GPT-3 requires substantial computation resources, motivating recent efforts to explore parameter-efficient adaptation to downstream tasks. One practical area of research is to treat these models as black boxes and interact with them through their inference APIs. In this paper, we investigate how to optimize few-shot text classification without accessing the gradients of the LLMs. To achieve this, we treat the black-box model as a feature extractor and train a classifier with the augmented text data. Data augmentation is performed using prompt-based finetuning on an auxiliary language model with a much smaller parameter size than the black-box model. Through extensive experiments on eight text classification datasets, we show that our approach, dubbed BT-Classifier, significantly outperforms state-of-the-art black-box few-shot learners and performs on par with methods that rely on full-model tuning.

cs.CL↗

arXiv2023

Spin-Statistics for Black Hole Microstates

The gravitational path integral can be used to compute the number of black hole states for a given energy window, or the free energy in a thermal ensemble. In this article we explain how to use the gravitational path integral to compute the separate number of bosonic and fermionic black hole microstates. We do this by comparing the partition function with and without the insertion of $(-1)^{\sf F}$. In particular we introduce a universal rotating black hole that contributes to the partition function in the presence of $(-1)^{\sf F}$. We study this problem for black holes in asymptotically flat space and in AdS, putting constraints on the high energy spectrum of holographic CFTs (not necessarily supersymmetric). Finally, we analyze wormhole contributions to related quantities.

hep-th↗

arXiv2023

Sharper Convergence Guarantees for Federated Learning with Partial Model Personalization

Partial model personalization, which encompasses both shared and personal variables in its formulation, is a critical optimization problem in federated learning. It balances individual client needs with collective knowledge utilization, and serves as a general formulation covering various key scenarios, ranging from fully shared to fully personalized federated learning. This paper introduces two effective algorithms, FedAvg-P and Scaffold-P, to solve this problem and provides sharp convergence analyses, quantifying the influence of gradient variance, local steps, and partial client sampling on their performance. Our established rates surpass existing results and, meanwhile, are based on more relaxed assumptions. Additionally, our analyses are also applicable to fully shared or fully personalized federated learning, matching or even outperforming their best known convergence rates. Numerical experiments corroborate our theoretical findings.

math.OC↗

arXiv2023

Influence of shear waves on transcranial ultrasound propagation in cortical brain regions

Transcranial ultrasound applications require accurate simulations to predict intracranial acoustic pressure fields. The current gold standard typically consists of calculating a longitudinal ultrasound wave propagation using a fluid skull model, which is based on full head CT images for retrieving the skull's geometry and elastic constants. Although this approach has extensively been validated for deep brain targets and routinely used in transcranial ultrasound ablation procedures, its accuracy in shallow cortical regions remains unexplored. In this study, we explore the shear wave effects associated with transcranial focused ultrasound propagation, both numerically and experimentally. The intracranial acoustic pressure was measured at different incidence angles at the parietal and frontal regions in an ex vivo human skull. The fluid-like skull model was then compared to the solid model comprising both longitudinal and shear waves. The results consistently show a larger error and variability for both models when considering an oblique incidence, reaching a maximum of 125% mean deviation of the focal area when employing the fluid skull model. Statistical assessments further revealed that ignoring shear waves results in an average ~40% overestimation of the intracranial acoustic pressure and inability to obtain an accurate intracranial acoustic pressure distribution. Moreover, the solid model has a more stable performance, even when small variations in the skull-transducer relative position are introduced. Our results could contribute to the refinement of the transcranial ultrasound propagation modeling methods thus help improving the safety and outcome of transcranial ultrasound therapy in the cortical brain areas.

physics.med-ph↗

arXiv2023

On Talagrand's functional and generic chaining

In the study of the supremum of stochastic processes, Talagrand's chaining functionals and his generic chaining method are heavily related to the distribution of stochastic processes. In the present paper, we construct Talagrand's type functionals in the general distribution case and obtain the upper bound for the suprema of all $p$-th moments of the stochastic process using the generic chaining method. As applications, we obtained the Johnson-Lindenstrauss lemma, the upper bound for the supremum of all $p$-th moment of order 2 Gaussian chaos, and convex signal recovery in our setting.

math.PR↗

arXiv2023

Probing Supersymmetric Black Holes with Surface Defects

It has long been conjectured that the large $N$ deconfinement phase transition of $\mathcal{N}=4$ ${\rm SU}(N)$ super-Yang-Mills corresponds via AdS/CFT to the Hawking-Page transition in which black holes dominate the thermal ensemble, and quantitative evidence of this has come through the recent matching of the superconformal index of ${1\over 16}$-BPS states to the supersymmetric black hole entropy. We introduce the half-BPS Gukov-Witten surface defect as a probe of the superconformal index, which also serves as an order parameter for the deconfinement transition. This can be studied directly in field theory as a modification of the usual unitary matrix model or in the dual description as a D3-brane probe in the background of a (complex) supersymmetric black hole. Using a saddle point approximation, we determine our defect index in the large $N$ limit as a simple function of the chemical potentials and show independently that it is reproduced by the renormalized action of the brane in the black hole background. Along the way, we also comment on the Cardy limit and the thermodynamics of the D3-brane in the generalized ensemble. The defect index sharply distinguishes between the confining and the deconfining phases of the gauge theory and thus is a supersymmetric non-perturbative order parameter for these large $N$ phase transitions which deserves further investigation. Finally, our work provides an example where the properties of a black hole coupled to an external system can be analyzed precisely.

hep-th↗

arXiv2023

Association of stroke lesion distributions with atrial fibrillation detected after stroke

Background Atrial fibrillation is often missed by traditional intermittent electrocardiogram monitoring after ischemic stroke due to its paroxysmal and asymptomatic nature. The knowledge of the unique characteristics of the population with atrial fibrillation detected after stroke (AFDAS) enables more ischemic stroke patients to benefit from more aggressive anticoagulation therapy and AF management. Method This is an observational, retrospective, MRI imaging-based single-center study. Patients with AFDAS were matched in 1:3 ratio with patients without AF (NoAF)and patients with known AF before stroke(KAF) in PSM model based on age, gender, and time from stroke onset to admission. Multivariate logistic models were used to test the association of MRI-based stroke lesion distribution, other clinical parameters and AF. A backward stepwise elimination regression was conducted to identify the most important variables. Results Compared to the NoAF group(n=103), the patients with AFDAS (n=42) had more cortical involvement(p=0.016), as well as temporal(p<0.001) and insular lobes(p=0.018) infraction. After performing a backward stepwise elimination model in regression analysis, the temporal lobe infraction(OR 0.274, 95%CI 0.090-0.838, p=0.023) remained independently associated with the detection of AF. Compared to the KAF group(n=89), LAD(OR 1.113, 95%CI 1.022-1.211, p=0.014), Number of lobes infarction(p=0.012), 3-lobes involvement(OR 0.177, 95%CI 0.056-0.559, p=0.003), and left hemisphere lobe involvement(OR 5.966, 95%CI 2.273-15.817, p<0.001)) were independently associated with AFDAS and KAF. Conclusions Ischemic stroke patients with AF detected after stroke present more temporal lobe infraction and cortical involvement. These lesion distribution characteristics with clinical characteristics together may help in stratifying patients with long-term cardiac monitoring after stroke.

q-bio.NC↗

arXiv2023

Dynamic Transformers Provide a False Sense of Efficiency

Despite much success in natural language processing (NLP), pre-trained language models typically lead to a high computational cost during inference. Multi-exit is a mainstream approach to address this issue by making a trade-off between efficiency and accuracy, where the saving of computation comes from an early exit. However, whether such saving from early-exiting is robust remains unknown. Motivated by this, we first show that directly adapting existing adversarial attack approaches targeting model accuracy cannot significantly reduce inference efficiency. To this end, we propose a simple yet effective attacking framework, SAME, a novel slowdown attack framework on multi-exit models, which is specially tailored to reduce the efficiency of the multi-exit models. By leveraging the multi-exit models' design characteristics, we utilize all internal predictions to guide the adversarial sample generation instead of merely considering the final prediction. Experiments on the GLUE benchmark show that SAME can effectively diminish the efficiency gain of various multi-exit models by 80% on average, convincingly validating its effectiveness and generalization ability.

cs.CL↗

arXiv2023

Low-intensity pulsed ultrasound promotes mesenchymal stem cell transplantation-based articular cartilage regeneration via inhibiting the TNF signaling pathway

Background: Mesenchymal stem cell (MSC) transplantation therapy is highly investigated for the regenerative repair of cartilage defects. Low-intensity pulsed ultrasound (LIPUS) has the potential to promote chondrogenic differentiation of MSCs. However, its underlying mechanism remains unclear. Here, we investigated the promoting effects and mechanisms underlying LIPUS stimulation on the chondrogenic differentiation of human umbilical cord mesenchymal stem cells (hUC-MSCs) and further evaluated its regenerative application value in articular cartilage defects in rats. Methods: LIPUS was applied to stimulate cultured hUC-MSCs and C28/I2 cells in vitro. Immunofluorescence staining, qPCR analysis, and transcriptome sequencing were used to detect mature cartilage-related markers of gene and protein expression for a comprehensive evaluation of differentiation. Injured articular cartilage rat models were established for further hUC-MSC transplantation and LIPUS stimulation in vivo. Histopathology and H&E staining were used to evaluate the repair effects of the injured articular cartilage with LIPUS stimulation. Results: The results showed that LIPUS stimulation with specific parameters effectively promoted the expression of mature cartilage-related genes and proteins, inhibited TNF-α gene expression in hUC-MSCs, and exhibited anti-inflammation in C28/I2 cells. In addition, the articular cartilage defects of rats were significantly repaired after hUC-MSC transplantation and LIPUS stimulation. Conclusions: Taken together, LIPUS stimulation could realize articular cartilage regeneration based on hUC-MSC transplantation due to the inhibition of the TNF signaling pathway, which is of clinical value for the relief of osteoarthritis.

physics.med-ph↗

arXiv2023

Effective Ambiguity Attack Against Passport-based DNN Intellectual Property Protection Schemes through Fully Connected Layer Substitution

Since training a deep neural network (DNN) is costly, the well-trained deep models can be regarded as valuable intellectual property (IP) assets. The IP protection associated with deep models has been receiving increasing attentions in recent years. Passport-based method, which replaces normalization layers with passport layers, has been one of the few protection solutions that are claimed to be secure against advanced attacks. In this work, we tackle the issue of evaluating the security of passport-based IP protection methods. We propose a novel and effective ambiguity attack against passport-based method, capable of successfully forging multiple valid passports with a small training dataset. This is accomplished by inserting a specially designed accessory block ahead of the passport parameters. Using less than 10% of training data, with the forged passport, the model exhibits almost indistinguishable performance difference (less than 2%) compared with that of the authorized passport. In addition, it is shown that our attack strategy can be readily generalized to attack other IP protection methods based on watermark embedding. Directions for potential remedy solutions are also given.

cs.CR↗

arXiv2023

Estimating global charge violating amplitudes from wormholes

We consider the scattering of high energy and ultra relativistic spherically symmetric shells in asymptotically AdS$_D$ spacetimes. We analyze an exclusive amplitude where a single spherically symmetric shell goes in and a single one comes out, such that the two have different global symmetry charges of the effective gravity theory. We study a simple wormhole configuration that computes the square of the amplitude and analyze its properties.

hep-th↗

arXiv2022

SEVGGNet-LSTM: a fused deep learning model for ECG classification

This paper presents a fused deep learning algorithm for ECG classification. It takes advantages of the combined convolutional and recurrent neural network for ECG classification, and the weight allocation capability of attention mechanism. The input ECG signals are firstly segmented and normalized, and then fed into the combined VGG and LSTM network for feature extraction and classification. An attention mechanism (SE block) is embedded into the core network for increasing the weight of important features. Two databases from different sources and devices are employed for performance validation, and the results well demonstrate the effectiveness and robustness of the proposed algorithm.

cs.LG↗

arXiv2022

Generate, Discriminate and Contrast: A Semi-Supervised Sentence Representation Learning Framework

Most sentence embedding techniques heavily rely on expensive human-annotated sentence pairs as the supervised signals. Despite the use of large-scale unlabeled data, the performance of unsupervised methods typically lags far behind that of the supervised counterparts in most downstream tasks. In this work, we propose a semi-supervised sentence embedding framework, GenSE, that effectively leverages large-scale unlabeled data. Our method include three parts: 1) Generate: A generator/discriminator model is jointly trained to synthesize sentence pairs from open-domain unlabeled corpus; 2) Discriminate: Noisy sentence pairs are filtered out by the discriminator to acquire high-quality positive and negative sentence pairs; 3) Contrast: A prompt-based contrastive approach is presented for sentence representation learning with both annotated and synthesized data. Comprehensive experiments show that GenSE achieves an average correlation score of 85.19 on the STS datasets and consistent performance improvement on four domain adaptation tasks, significantly surpassing the state-of-the-art methods and convincingly corroborating its effectiveness and generalization ability.Code, Synthetic data and Models available at https://github.com/MatthewCYM/GenSE.

cs.CL↗