arXiv subjects

Yi Ding

Publications and source records attributed to Yi Ding.

At least 145 records · Page 8Linked to original sources

arXiv2023

Generalizability of PRS313 for breast cancer risk amongst non-Europeans in a Los Angeles biobank

Polygenic risk scores (PRS) summarize the combined effect of common risk variants and are associated with breast cancer risk in patients without identifiable monogenic risk factors. One of the most well-validated PRSs in breast cancer to date is PRS313, which was developed from a Northern European biobank but has shown attenuated performance in non-European ancestries. We further investigate the generalizability of the PRS313 for American women of European (EA), African (AFR), Asian (EAA), and Latinx (HL) ancestry within one institution with a singular EHR system, genotyping platform, and quality control process. We found that the PRS313 achieved overlapping Areas under the ROC Curve (AUCs) in females of Lantix (AUC, 0.68; 95 CI, 0.65-0.71) and European ancestry (AUC, 0.70; 95 CI, 0.69-0.71) but lower AUCs for the AFR and EAA populations (AFR: AUC, 0.61; 95 CI, 0.56-0.65; EAA: AUC, 0.64; 95 CI, 0.60-0.680). While PRS313 is associated with Hormone Positive (HR+) disease in European Americans (OR, 1.42; 95 CI, 1.16-1.64), for Latinx females, it may be instead associated with Human Epidermal Growth Factor Receptor 2 (HER2+) disease (OR, 2.52; 95 CI, 1.35-4.70) although due to small numbers, additional studies are needed. In summary, we found that PRS313 was significantly associated with breast cancer but with attenuated accuracy in women of African and Asian descent within a singular health system in Los Angeles. Our work further highlights the need for additional validation in diverse cohorts prior to clinical implementation of polygenic risk scores.

q-bio.GN↗

arXiv2022

Subsampling Spectral Clustering for Large-Scale Social Networks

Online social network platforms such as Twitter and Sina Weibo have been extremely popular over the past 20 years. Identifying the network community of a social platform is essential to exploring and understanding the users' interests. However, the rapid development of science and technology has generated large amounts of social network data, creating great computational challenges for community detection in large-scale social networks. Here, we propose a novel subsampling spectral clustering algorithm to identify community structures in large-scale social networks with limited computing resources. More precisely, spectral clustering is conducted using only the information of a small subsample of the network nodes, resulting in a huge reduction in computational time. As a result, for large-scale datasets, the method can be realized even using a personal computer. Specifically, we introduce two different sampling techniques, namely simple random subsampling and degree corrected subsampling. The methodology is applied to the dataset collected from Sina Weibo, which is one of the largest Twitter-type social network platforms in China. Our method can very effectively identify the community structure of registered users. This community structure information can be applied to help Sina Weibo promote advertisements to target users and increase user activity.

cs.SI↗

arXiv2022

Fast DistilBERT on CPUs

Transformer-based language models have become the standard approach to solving natural language processing tasks. However, industry adoption usually requires the maximum throughput to comply with certain latency constraints that prevents Transformer models from being used in production. To address this gap, model compression techniques such as quantization and pruning may be used to improve inference efficiency. However, these compression techniques require specialized software to apply and deploy at scale. In this work, we propose a new pipeline for creating and running Fast Transformer models on CPUs, utilizing hardware-aware pruning, knowledge distillation, quantization, and our own Transformer inference runtime engine with optimized kernels for sparse and quantized operators. We demonstrate the efficiency of our pipeline by creating a Fast DistilBERT model showing minimal accuracy loss on the question-answering SQuADv1.1 benchmark, and throughput results under typical production constraints and environments. Our results outperform existing state-of-the-art Neural Magic's DeepSparse runtime performance by up to 50% and up to 4.1x performance speedup over ONNX Runtime. Source code is publicly available at https://github.com/intel/intel-extension-for-transformers.

cs.CL↗

arXiv2022

LGGNet: Learning from Local-Global-Graph Representations for Brain-Computer Interface

Neuropsychological studies suggest that co-operative activities among different brain functional areas drive high-level cognitive processes. To learn the brain activities within and among different functional areas of the brain, we propose LGGNet, a novel neurologically inspired graph neural network, to learn local-global-graph representations of electroencephalography (EEG) for Brain-Computer Interface (BCI). The input layer of LGGNet comprises a series of temporal convolutions with multi-scale 1D convolutional kernels and kernel-level attentive fusion. It captures temporal dynamics of EEG which then serves as input to the proposed local and global graph-filtering layers. Using a defined neurophysiologically meaningful set of local and global graphs, LGGNet models the complex relations within and among functional areas of the brain. Under the robust nested cross-validation settings, the proposed method is evaluated on three publicly available datasets for four types of cognitive classification tasks, namely, the attention, fatigue, emotion, and preference classification tasks. LGGNet is compared with state-of-the-art methods, such as DeepConvNet, EEGNet, R2G-STNN, TSception, RGNN, AMCNN-DGCN, HRNN and GraphNet. The results show that LGGNet outperforms these methods, and the improvements are statistically significant (p<0.05) in most cases. The results show that bringing neuroscience prior knowledge into neural network design yields an improvement of classification performance. The source code can be found at https://github.com/yi-ding-cs/LGG

cs.NE↗

arXiv2022

High-dimensional covariance matrices under dynamic volatility models: asymptotics and shrinkage estimation

We study the estimation of the high-dimensional covariance matrix andits eigenvalues under dynamic volatility models. Data under such modelshave nonlinear dependency both cross-sectionally and temporally. We firstinvestigate the empirical spectral distribution (ESD) of the sample covariancematrix under scalar BEKK models and establish conditions under which thelimiting spectral distribution (LSD) is either the same as or different fromthe i.i.d. case. We then propose a time-variation adjusted (TV-adj) sample co-variance matrix and prove that its LSD follows the same Marcenko-Pasturlaw as the i.i.d. case. Based on the asymptotics of the TV-adj sample co-variance matrix, we develop a consistent population spectrum estimator and an asymptotically optimal nonlinear shrinkage estimator of the unconditionalcovariance matrix

math.ST↗

arXiv2022

Statistical Learning for Individualized Asset Allocation

We establish a high-dimensional statistical learning framework for individualized asset allocation. Our proposed methodology addresses continuous-action decision-making with a large number of characteristics. We develop a discretization approach to model the effect of continuous actions and allow the discretization frequency to be large and diverge with the number of observations. The value function of continuous-action is estimated using penalized regression with our proposed generalized penalties that are imposed on linear transformations of the model coefficients. We show that our proposed Discretization and Regression with generalized fOlded concaVe penalty on Effect discontinuity (DROVE) approach enjoys desirable theoretical properties and allows for statistical inference of the optimal value associated with optimal decision-making. Empirically, the proposed framework is exercised with the Health and Retirement Study data in finding individualized optimal asset allocation. The results show that our individualized optimal strategy improves the population financial well-being.

stat.ML↗

arXiv2022

NURD: Negative-Unlabeled Learning for Online Datacenter Straggler Prediction

Datacenters execute large computational jobs, which are composed of smaller tasks. A job completes when all its tasks finish, so stragglers -- rare, yet extremely slow tasks -- are a major impediment to datacenter performance. Accurately predicting stragglers would enable proactive intervention, allowing datacenter operators to mitigate stragglers before they delay a job. While much prior work applies machine learning to predict computer system performance, these approaches rely on complete labels -- i.e., sufficient examples of all possible behaviors, including straggling and non-straggling -- or strong assumptions about the underlying latency distributions -- e.g., whether Gaussian or not. Within a running job, however, none of this information is available until stragglers have revealed themselves when they have already delayed the job. To predict stragglers accurately and early without labeled positive examples or assumptions on latency distributions, this paper presents NURD, a novel Negative-Unlabeled learning approach with Reweighting and Distribution-compensation that only trains on negative and unlabeled streaming data. The key idea is to train a predictor using finished tasks of non-stragglers to predict latency for unlabeled running tasks, and then reweight each unlabeled task's prediction based on a weighting function of its feature space. We evaluate NURD on two production traces from Google and Alibaba, and find that compared to the best baseline approach, NURD produces 2--11 percentage point increases in the F1 score in terms of prediction accuracy, and 2.0--8.8 percentage point improvements in job completion time.

cs.LG↗

arXiv2022

Hanbury Brown-Twiss effect with electromagnetic scattered field generated by a collection of particles of $\mathcal{L}$ types

A theoretical framework in the spherical polar coordinate system is developed to systematically treat the correlation between intensity fluctuations (CIF) of electromagnetic light waves on scattering from a collection of particles of $\mathcal{L}$ types. Two $\mathcal{L}\times \mathcal{L}$ matrices called pair-potential matrix (PPM) and pair-structure matrix (PSM) are introduced to jointly formulate the normalized CIF of the scattered field for the first time. We build a closed-form relation that associates the normalized CIF with the PPM and the PSM as well as the spectral degree of polarization $\mathcal{P}$ of the incident field, showing that the normalized CIF is closely related to the trace of the product of the PSM and the transpose of the PPM, and its dependence on $\mathcal{P}$ is completely determined by the scattering polar angle and azimuth angle. For a special case where the spatial distributions of scattering potentials of particles of different types are similar and the same is true of their density distributions, the PPM and the PSM will reduce to two new matrices whose elements separately quantify the degree of angular correlation of the scattering potentials of particles and their density distributions, and the number of species of particles in this special case appears as a scaled factor to ensure the normalization of the CIF. The effects of the off-diagonal elements of the PPM and the PSM on the normalized CIF and its dependence on $\mathcal{P}$ are illustrated by two numerical examples.

physics.optics↗

arXiv2022

Scattering of light waves from a collection with particles of $\mathcal{L}$ types

A new approach is developed within the first-order Born approximation to light scattering from a collection of particles with $\mathcal{L}$ types. Two $\mathcal{L}\times\mathcal{L}$ matrices called pair-potential matrix (PPM) and pair-structure matrix (PSM) are introduced to jointly formulate the coherence properties of the scattered field. We derive a closed-form relation that associates the cross-spectral density function of the scattered field with the PPM and the PSM, showing that the the cross-spectral density function equals the trace of the product of the PSM and the transpose of the PPM. Based on this, the spectral degree of coherence (SDOC) of the scattered field is further analysed. We show that for a special case where the spatial distributions of scattering potentials of different types of particles are similar and the same is true of their density distributions, the PPM and the PSM will reduce to two new matrices whose elements separately quantify degree of angular correlation of the scattering potentials of particles and their density distributions, and the number of species of particles in this special case as a scaled factor ensures the normalization of the SDOC. Two special hybrid particulate systems as examples are given to illustrate the importance of our new approach.

physics.optics↗

arXiv2022

TSception: Capturing Temporal Dynamics and Spatial Asymmetry from EEG for Emotion Recognition

The high temporal resolution and the asymmetric spatial activations are essential attributes of electroencephalogram (EEG) underlying emotional processes in the brain. To learn the temporal dynamics and spatial asymmetry of EEG towards accurate and generalized emotion recognition, we propose TSception, a multi-scale convolutional neural network that can classify emotions from EEG. TSception consists of dynamic temporal, asymmetric spatial, and high-level fusion layers, which learn discriminative representations in the time and channel dimensions simultaneously. The dynamic temporal layer consists of multi-scale 1D convolutional kernels whose lengths are related to the sampling rate of EEG, which learns the dynamic temporal and frequency representations of EEG. The asymmetric spatial layer takes advantage of the asymmetric EEG patterns for emotion, learning the discriminative global and hemisphere representations. The learned spatial representations will be fused by a high-level fusion layer. Using more generalized cross-validation settings, the proposed method is evaluated on two publicly available datasets DEAP and MAHNOB-HCI. The performance of the proposed network is compared with prior reported methods such as SVM, KNN, FBFgMDM, FBTSC, Unsupervised learning, DeepConvNet, ShallowConvNet, and EEGNet. TSception achieves higher classification accuracies and F1 scores than other methods in most of the experiments. The codes are available at https://github.com/yi-ding-cs/TSception

cs.LG↗

arXiv2022

SCOPE: Safe Exploration for Dynamic Computer Systems Optimization

Modern computer systems need to execute under strict safety constraints (e.g., a power limit), but doing so often conflicts with their ability to deliver high performance (i.e. minimal latency). Prior work uses machine learning to automatically tune hardware resources such that the system execution meets safety constraints optimally. Such solutions monitor past system executions to learn the system's behavior under different hardware resource allocations before dynamically tuning resources to optimize the application execution. However, system behavior can change significantly between different applications and even different inputs of the same applications. Hence, the models learned using data collected a priori are often suboptimal and violate safety constraints when used with new applications and inputs. To address this limitation, we introduce the concept of an execution space, which is the cross product of hardware resources, input features, and applications. To dynamically and safely allocate hardware resources from the execution space, we present SCOPE, a resource manager that leverages a novel safe exploration framework. We evaluate SCOPE's ability to deliver improved latency while minimizing power constraint violations by dynamically configuring hardware while running a variety of Apache Spark applications. Compared to prior approaches that minimize power constraint violations, SCOPE consumes comparable power while improving latency by up to 9.5X. Compared to prior approaches that minimize latency, SCOPE achieves similar latency but reduces power constraint violation rates by up to 45.88X, achieving almost zero safety constraint violations across all applications.

eess.SY↗

arXiv2022

Cello: Efficient Computer Systems Optimization with Predictive Early Termination and Censored Regression

Sample-efficient machine learning (SEML) has been widely applied to find optimal latency and power tradeoffs for configurable computer systems. Instead of randomly sampling from the configuration space, SEML reduces the search cost by dramatically reducing the number of configurations that must be sampled to optimize system goals (e.g., low latency or energy). Nevertheless, SEML only reduces one component of cost -- the total number of samples collected -- but does not decrease the cost of collecting each sample. Critically, not all samples are equal; some take much longer to collect because they correspond to slow system configurations. This paper present Cello, a computer systems optimization framework that reduces sample collection costs -- especially those that come from the slowest configurations. The key insight is to predict ahead of time whether samples will have poor system behavior (e.g., long latency or high energy) and terminate these samples early before their measured system behavior surpasses the termination threshold, which we call it predictive early termination. To predict the future system behavior accurately before it manifests as high runtime or energy, Cello uses censored regression to produces accurate predictions for running samples. We evaluate Cello by optimizing latency and energy for Apache Spark workloads. We give Cello a fixed amount of time to search a combined space of hardware and software configuration parameters. Our evaluation shows that compared to the state-of-the-art SEML approach in computer systems optimization, Cello improves latency by 1.19X for minimizing latency under a power constraint, and improves energy by 1.18X for minimizing energy under a latency constraint.

cs.LG↗

arXiv2022

Continuous Emotion Recognition using Visual-audio-linguistic information: A Technical Report for ABAW3

We propose a cross-modal co-attention model for continuous emotion recognition using visual-audio-linguistic information. The model consists of four blocks. The visual, audio, and linguistic blocks are used to learn the spatial-temporal features of the multi-modal input. A co-attention block is designed to fuse the learned features with the multi-head co-attention mechanism. The visual encoding from the visual block is concatenated with the attention feature to emphasize the visual information. To make full use of the data and alleviate over-fitting, cross-validation is carried out on the training and validation set. The concordance correlation coefficient (CCC) centering is used to merge the results from each fold. The achieved CCC on the test set is $0.520$ for valence and $0.602$ for arousal, which significantly outperforms the baseline method with the corresponding CCC of 0.180 and 0.170 for valence and arousal, respectively. The code is available at https://github.com/sucv/ABAW3.

cs.MM↗

arXiv2021

Programming with Neural Surrogates of Programs

Surrogates, models that mimic the behavior of programs, form the basis of a variety of development workflows. We study three surrogate-based design patterns, evaluating each in case studies on a large-scale CPU simulator. With surrogate compilation, programmers develop a surrogate that mimics the behavior of a program to deploy to end-users in place of the original program. Surrogate compilation accelerates the CPU simulator under study by $1.6\times$. With surrogate adaptation, programmers develop a surrogate of a program then retrain that surrogate on a different task. Surrogate adaptation decreases the simulator's error by up to $50\%$. With surrogate optimization, programmers develop a surrogate of a program, optimize input parameters of the surrogate, then plug the optimized input parameters back into the original program. Surrogate optimization finds simulation parameters that decrease the simulator's error by $5\%$ compared to the error induced by expert-set parameters. In this paper we formalize this taxonomy of surrogate-based design patterns. We further describe the programming methodology common to all three design patterns. Our work builds a foundation for the emerging class of workflows based on programming with surrogates of programs.

cs.PL↗

arXiv2021

Sparse Fusion for Multimodal Transformers

Multimodal classification is a core task in human-centric machine learning. We observe that information is highly complementary across modalities, thus unimodal information can be drastically sparsified prior to multimodal fusion without loss of accuracy. To this end, we present Sparse Fusion Transformers (SFT), a novel multimodal fusion method for transformers that performs comparably to existing state-of-the-art methods while having greatly reduced memory footprint and computation cost. Key to our idea is a sparse-pooling block that reduces unimodal token sets prior to cross-modality modeling. Evaluations are conducted on multiple multimodal benchmark datasets for a wide range of classification tasks. State-of-the-art performance is obtained on multiple benchmarks under similar experiment conditions, while reporting up to six-fold reduction in computational cost and memory requirements. Extensive ablation studies showcase our benefits of combining sparsification and multimodal learning over naive approaches. This paves the way for enabling multimodal learning on low-resource devices.

cs.CV↗

arXiv2021

Continuous Emotion Recognition with Audio-visual Leader-follower Attentive Fusion

We propose an audio-visual spatial-temporal deep neural network with: (1) a visual block containing a pretrained 2D-CNN followed by a temporal convolutional network (TCN); (2) an aural block containing several parallel TCNs; and (3) a leader-follower attentive fusion block combining the audio-visual information. The TCN with large history coverage enables our model to exploit spatial-temporal information within a much larger window length (i.e., 300) than that from the baseline and state-of-the-art methods (i.e., 36 or 48). The fusion block emphasizes the visual modality while exploits the noisy aural modality using the inter-modality attention mechanism. To make full use of the data and alleviate over-fitting, cross-validation is carried out on the training and validation set. The concordance correlation coefficient (CCC) centering is used to merge the results from each fold. On the test (validation) set of the Aff-Wild2 database, the achieved CCC is 0.463 (0.469) for valence and 0.492 (0.649) for arousal, which significantly outperforms the baseline method with the corresponding CCC of 0.200 (0.210) and 0.190 (0.230) for valence and arousal, respectively. The code is available at https://github.com/sucv/ABAW2.

cs.CV↗

arXiv2021

Channel Pruning Guided by Spatial and Channel Attention for DNNs in Intelligent Edge Computing

Deep Neural Networks (DNNs) have achieved remarkable success in many computer vision tasks recently, but the huge number of parameters and the high computation overhead hinder their deployments on resource-constrained edge devices. It is worth noting that channel pruning is an effective approach for compressing DNN models. A critical challenge is to determine which channels are to be removed, so that the model accuracy will not be negatively affected. In this paper, we first propose Spatial and Channel Attention (SCA), a new attention module combining both spatial and channel attention that respectively focuses on "where" and "what" are the most informative parts. Guided by the scale values generated by SCA for measuring channel importance, we further propose a new channel pruning approach called Channel Pruning guided by Spatial and Channel Attention (CPSCA). Experimental results indicate that SCA achieves the best inference accuracy, while incurring negligibly extra resource consumption, compared to other state-of-the-art attention modules. Our evaluation on two benchmark datasets shows that, with the guidance of SCA, our CPSCA approach achieves higher inference accuracy than other state-of-the-art pruning methods under the same pruning ratios.

cs.CV↗

arXiv2021

Augmentation Strategies for Learning with Noisy Labels

Imperfect labels are ubiquitous in real-world datasets. Several recent successful methods for training deep neural networks (DNNs) robust to label noise have used two primary techniques: filtering samples based on loss during a warm-up phase to curate an initial set of cleanly labeled samples, and using the output of a network as a pseudo-label for subsequent loss calculations. In this paper, we evaluate different augmentation strategies for algorithms tackling the "learning with noisy labels" problem. We propose and examine multiple augmentation strategies and evaluate them using synthetic datasets based on CIFAR-10 and CIFAR-100, as well as on the real-world dataset Clothing1M. Due to several commonalities in these algorithms, we find that using one set of augmentations for loss modeling tasks and another set for learning is the most effective, improving results on the state-of-the-art and other previous methods. Furthermore, we find that applying augmentation during the warm-up period can negatively impact the loss convergence behavior of correctly versus incorrectly labeled samples. We introduce this augmentation strategy to the state-of-the-art technique and demonstrate that we can improve performance across all evaluated noise levels. In particular, we improve accuracy on the CIFAR-10 benchmark at 90% symmetric noise by more than 15% in absolute accuracy, and we also improve performance on the Clothing1M dataset. (K. Nishi and Y. Ding contributed equally to this work)

cs.CV↗