This paper presents a novel approach to the joint optimization of job scheduling and data allocation in grid computing environments. We formulate this joint optimization problem as a mixed integer quadratically constrained program. To tackle the nonlinearity in the constraint, we alternatively fix a subset of decision variables and optimize the remaining ones via Mixed Integer Linear Programming (MILP). We solve the MILP problem at each iteration via an off-the-shelf MILP solver. Our experimental results show that our method significantly outperforms existing heuristic methods, employing either independent optimization or joint optimization strategies. We have also verified the generalization ability of our method over grid environments with various sizes and its high robustness to the algorithm setting.

OBJECTIVE: Phase II of MVP-CHAMPION, a federal collaboration between the Veterans Affairs Healthcare System (VA) and the Department of Energy (DoE), leveraged large-scale clinical, geo-spatial, and genetic data with state-of-the-art artificial intelligence (AI), and high-performance computing (HPC) to improve value in healthcare. MATERIALS AND METHODS: Eight clinical priority projects for which AI was a critical missing capability were initiated to address: lung cancer screening (MVP 061), suicide risk screening (MVP 062), cardiovascular risk in obstructive sleep apnea (MVP 063), checkpoint inhibitor toxicity (MVP 064), heart failure (MVP 065), renal complications in diabetes (MVP 066), post COVID-19 sequelae (MVP 067), and antipsychotic medication toxicity (MVP 068). RESULTS: Building on a strong regulatory and administrative foundation, we developed multimorbidity-aware analytic frameworks, reusable computational tools, and analytic pipelines. These greatly facilitated identification of novel risk factors including genetic variants and specification of more discriminating prediction models. Novel genetic risk factors are informing development and repurposing of medications and discriminating prediction models promise to improve healthcare value. DISCUSSION: The research foundation developed in Phase I and extended in Phase II of MVP CHAMPION has supported an unprecedented federal collaboration and yielded significant scientific advances. Our clinical findings are poised for near-term application, while advances in machine learning and high-performance computing may accelerate the broader adoption of artificial intelligence in healthcare. CONCLUSION: This maturing VA-DoE federal collaboration is poised to transform the future of Veterans' healthcare and the broader national landscape of precision health.

Particle accelerators are composed of various components, and their properties are finely tuned to optimize certain particle beam qualities as they accelerate. In particular, particle colliders like the Relativistic Heavy Ion Collider (RHIC) at Brookhaven National Lab (BNL) are interested in maximizing luminosity, a measure of the collision rate primarily determined by the beam intensity (number of particles) and its beam size. However, finding and maintaining optimum settings is a time-consuming expert operator activity. This work proposes the use of the Recurrent Proximal Policy Optimization (RPPO), a reinforcement learning algorithm, to find parameters of quadrupole magnet strengths optimizing the beam qualities in the Booster to AGS (Alternating Gradient Synchrotron) section of the RHIC complex.

Natural language prompt optimization, or prompt engineering, has emerged as a powerful technique to unlock the potential of Large Language Models (LLMs) for various tasks. While existing methods primarily focus on maximizing a single task-specific performance metric for LLM outputs, real-world applications often require considering trade-offs between multiple objectives. In this work, we address this limitation by proposing an effective technique for multi-objective prompt optimization for LLMs. Specifically, we propose **ParetoPrompt**, a reinforcement learning~(RL) method that leverages dominance relationships between prompts to derive a policy model for prompts optimization using preference-based loss functions. By leveraging multi-objective dominance relationships, ParetoPrompt enables efficient exploration of the entire Pareto front without the need for a predefined scalarization of multiple objectives. Our experimental results show that ParetoPrompt consistently outperforms existing algorithms that use specific objective values. ParetoPrompt also yields robust performances when the objective metrics differ between training and testing.

Fourier Neural Operators (FNOs) offer a principled approach for solving complex partial differential equations (PDEs). However, scaling them to handle more complex PDEs requires increasing the number of Fourier modes, which significantly expands the number of model parameters and makes hyperparameter tuning computationally impractical. To address this, we introduce $\mu$Transfer-FNO, a zero-shot hyperparameter transfer technique that enables optimal configurations, tuned on smaller FNOs, to be directly applied to billion-parameter FNOs without additional tuning. Building on the Maximal Update Parametrization ($\mu$P) framework, we mathematically derive a parametrization scheme that facilitates the transfer of optimal hyperparameters across models with different numbers of Fourier modes in FNOs, which is validated through extensive experiments on various PDEs. Our empirical study shows that $\mu$Transfer-FNO reduces computational cost for tuning hyperparameters on large FNOs while maintaining or improving accuracy.

Spatiotemporal learning is challenging due to the intricate interplay between spatial and temporal dependencies, the high dimensionality of the data, and scalability constraints. These challenges are further amplified in scientific domains, where data is often irregularly distributed (e.g., missing values from sensor failures) and high-volume (e.g., high-fidelity simulations), posing additional computational and modeling difficulties. In this paper, we present SCENT, a novel framework for scalable and continuity-informed spatiotemporal representation learning. SCENT unifies interpolation, reconstruction, and forecasting within a single architecture. Built on a transformer-based encoder-processor-decoder backbone, SCENT introduces learnable queries to enhance generalization and a query-wise cross-attention mechanism to effectively capture multi-scale dependencies. To ensure scalability in both data size and model complexity, we incorporate a sparse attention mechanism, enabling flexible output representations and efficient evaluation at arbitrary resolutions. We validate SCENT through extensive simulations and real-world experiments, demonstrating state-of-the-art performance across multiple challenging tasks while achieving superior scalability.

Fitting Gaussian Processes (GPs) provides interpretable aleatoric uncertainty quantification for estimation of spatio-temporal fields. Spatio-temporal deep learning models, while scalable, typically assume a simplistic independent covariance matrix for the response, failing to capture the underlying correlation structure. However, spatio-temporal GPs suffer from issues of scalability and various forms of approximation bias resulting from restrictive assumptions of the covariance kernel function. We propose STACI, a novel framework consisting of a variational Bayesian neural network approximation of non-stationary spatio-temporal GP along with a novel spatio-temporal conformal inference algorithm. STACI is highly scalable, taking advantage of GPU training capabilities for neural network models, and provides statistically valid prediction intervals for uncertainty quantification. STACI outperforms competing GPs and deep methods in accurately approximating spatio-temporal processes and we show it easily scales to datasets with millions of observations.

Estimating causal effects from spatiotemporal observational data is essential in public health, environmental science, and policy evaluation, where randomized experiments are often infeasible. Existing approaches, however, either rely on strong structural assumptions or fail to handle key challenges such as interference, spatial confounding, temporal carryover, and *time-varying confounding*—where covariates are influenced by past treatments and, in turn, affect future ones. We introduce the **GST-UNet** (**G**-computation **S**patio-**T**emporal **UNet**), a theoretically grounded neural framework that combines a U-Net-based spatiotemporal encoder with regression-based iterative G-computation to estimate location-specific potential outcomes under complex intervention sequences. GST-UNet explicitly adjusts for time-varying confounders and captures non-linear spatial and temporal dependencies, enabling valid causal inference from a *single* observed trajectory in data-scarce settings. We validate its effectiveness in synthetic experiments and in a real-world analysis of wildfire smoke exposure and respiratory hospitalizations during the 2018 California Camp Fire. Together, these results position GST-UNet as a **principled and ready-to-use framework** for spatiotemporal causal inference, advancing reliable estimation in policy-relevant and scientific domains.

Efficiently processing time series data streams in real-time on resource-constrained devices offers significant advantages in terms of enhanced computational energy efficiency and reduced time-related risks. We introduce an innovative streaming time series classification network that utilizes attentive power iteration, enabling real-time processing on resource-constrained devices. Our model continuously updates a compact representation of the entire time series, enhancing classification accuracy while conserving energy and processing time. Notably, it excels in streaming scenarios without requiring complete time series access, enabling swift decisions. Experimental results show that our approach excels in classification accuracy and energy efficiency, with over 70% less consumption and threefold faster task completion than benchmarks. This work advances real-time responsiveness, energy conservation, and operational effectiveness for constrained devices, contributing to optimizing various applications.

Neural style transfer (NST) has evolved significantly in recent years. Yet, despite its rapid progress and advancement, existing NST methods either struggle to transfer aesthetic information from a style effectively or suffer from high computational costs and inefficiencies in feature disentanglement due to using pre-trained models. This work proposes a lightweight but effective model, AesFA---Aesthetic Feature-Aware NST. The primary idea is to decompose the image via its frequencies to better disentangle aesthetic styles from the reference image while training the entire model in an end-to-end manner to exclude pre-trained models at inference completely. To improve the network's ability to extract more distinct representations and further enhance the stylization quality, this work introduces a new aesthetic feature: contrastive loss. Extensive experiments and ablations show the approach not only outperforms recent NST methods in terms of stylization quality, but it also achieves faster inference. Codes are available at https://github.com/Sooyyoungg/AesFA.

The preservation of privacy is a critical concern in the implementation of artificial intelligence on sensitive training data. There are several techniques to preserve data privacy but quantum computations are inherently more secure due to the nocloning theorem, resulting in a most desirable computational platform on top of the potential quantum advantages. There have been prior works in protecting data privacy by Quantum Federated Learning (QFL) and Quantum Differential Privacy (QDP) studied independently. However, to the best of our knowledge, no prior work has addressed both QFL and QDP together yet. Here, we propose to combine these privacy-preserving methods and implement them on the quantum platform, so that we can achieve comprehensive protection against data leakage (QFL) and model inversion attacks (QDP). This implementation promises more efficient and secure artificial intelligence. In this paper, we present a successful implementation of these privacy-preservation methods by performing the binary classification of the Cats vs Dogs dataset. Using our quantum-classical machine learning model, we obtained a test accuracy of over 0.98, while maintaining epsilon values less than 1.3. We show that federated differentially private training is a viable privacy preservation method for quantum machine learning on Noisy Intermediate-Scale Quantum (NISQ) devices.

A major concern of deep learning models is the large amount of data that is required to build and train them, much of which is reliant on sensitive and personally identifiable information that is vulnerable to access by third parties. Ideas of using the quantum internet to address this issue have been previously proposed, which would enable fast and completely secure online communications. Previous work has yielded a hybrid quantum-classical transfer learning scheme for classical data and communication with a hub-spoke topology. While quantum communication is secure from eavesdrop attacks and no measurements from quantum to classical translation, due to no cloning theorem, hub-spoke topology is not ideal for quantum communication without quantum memory. Here we seek to improve this model by implementing a decentralized ring topology for the federated learning scheme, where each client is given a portion of the entire dataset and only performs training on that set. We also demonstrate the first successful use of quantum weights for quantum federated learning, which allows us to perform our training entirely in quantum.

The utility of machine learning has rapidly expanded in the last two decades and presented an ethical challenge. Papernot et. al. developed a technique, known as Private Aggregation of Teacher Ensembles (PATE) to enable federated learning in which multiple distributed teachers are trained on disjoint data sets. This study is the first to apply PATE to an ensemble of quantum neural networks (QNN) to pave a new way of ensuring privacy in quantum machine learning (QML).

Reliably reconstructing physical fields from sparse sensor data is a challenge that frequenty arises in many scientific domains. In practice, the process generating the data is often not known to sufficient accuracy. Therefore, there is a growing interest in the deep neural network route to the problem. In this work, we present a novel approach that learns a continuous representation of the field using implicit neural representations (INR). Specifically, after factorizing spatiotemporal variability into spatial and temporal components using the technique of separation of variables, the method learns relevant basis functions from sparsely sampled irregular data points to thus develop a continuous representation of the data. In experimental evaluations, the proposed model outperforms recent INR methods, offering superior reconstruction quality on simulation data from a state of the art climate model and on a second dataset that comprises of ultra-high resolution satellite-based sea surface temperature field. [Website for the Project: Both data and code are accessible.](https://xihaier.github.io/ICLR-2024-MMGN/)

Computed tomography (CT) scans are widely used for diagnosing and monitoring renal diseases but have a misdiagnosis rate of nearly 30%. Neural networks, particularly when integrated with quantum architectures, can significantly reduce this rate. This study introduces a Hybrid Matrix Product State (MPS) and Variational Quantum Circuit (VQC) architecture, where MPS is used for feature extraction and VQC for classification. The MPS-VQC model demonstrated enhanced performance, achieving 99.49% accuracy in distinguishing between kidney cysts, tumors, and healthy scans, compared to 52.92% using Principal Component Analysis (PCA)-VQC. As medical imaging increasingly adopts machine learning, quantum-inspired architectures like MPS-VQC offer a promising solution for improving diagnostic accuracy in renal disease detection and can be adapted for broader medical imaging tasks to enhance clinical outcomes and efficiency.

Domain generalization (DG) is a fundamental yet challenging topic in machine learning. Recently, the remarkable zero-shot capabilities of the large pre-trained vision-language model (e.g., CLIP) have made it popular for various downstream tasks. However, the effectiveness of this capacity often degrades when there are shifts in data distribution during testing compared to the training data. In this paper, we propose a novel method, known as CLIPCEIL, a model that utilizes Channel rEfinement and Image-text aLignment to facilitate the CLIP to the inaccessible $\textit{out-of-distribution}$ test datasets that exhibit domain shifts. Specifically, we refine the feature channels in the visual domain to ensure they contain domain-invariant and class-relevant features by using a lightweight adapter. This is achieved by minimizing the inter-domain variance while maximizing the inter-class variance. In the meantime, we ensure the image-text alignment by aligning text embeddings of the class descriptions and their corresponding image embedding while further removing the domain-specific features. Moreover, our model integrates multi-scale CLIP features by utilizing a self-attention fusion module, technically implemented through one Transformer layer. Extensive experiments on five widely used benchmark datasets demonstrate that CLIPCEIL outperforms the existing state-of-the-art methods. The source code is available at \url{https://github.com/yuxi120407/CLIPCEIL}.

While deep neural networks (DNNs) have revolutionized many fields, their fragility to carefully designed adversarial attacks impedes the usage of DNNs in safety-critical applications. In this article, we strive to explore the robust features that are not affected by the adversarial perturbations, that is, invariant to the clean image and its adversarial examples (AEs), to improve the model's adversarial robustness. Specifically, we propose a feature disentanglement model to segregate the robust features from nonrobust features and domain-specific features. The extensive experiments on five widely used datasets with different attacks demonstrate that robust features obtained from our model improve the model's adversarial robustness compared to the state-of-the-art approaches. Moreover, the trained domain discriminator is able to identify the domain-specific features from the clean images and AEs almost perfectly. This enables AE detection without incurring additional computational costs. With that, we can also specify different classifiers for clean images and AEs, thereby avoiding any drop in clean image accuracy.

Domain Generalization (DG) aims to learn a generalizable model on the unseen target domain by only training on the multiple observed source domains. Although a variety of DG methods have focused on extracting domain-invariant features, the domain-specific class-relevant features have attracted attention and been argued to benefit generalization to the unseen target domain. To take into account the class-relevant domain-specific information, in this paper we propose an Information theory iNspired diSentanglement and pURification modEl (INSURE) to explicitly disentangle the latent features to obtain sufficient and compact (necessary) class-relevant feature for generalization to the unseen domain. Specifically, we first propose an information theory inspired loss function to ensure the disentangled class-relevant features contain sufficient class label information and the other disentangled auxiliary feature has sufficient domain information. We further propose a paired purification loss function to let the auxiliary feature discard all the class-relevant information and thus the class-relevant feature will contain sufficient and compact (necessary) class-relevant information. Moreover, instead of using multiple encoders, we propose to use a learnable binary mask as our disentangler to make the disentanglement more efficient and make the disentangled features complementary to each other. We conduct extensive experiments on five widely used DG benchmark datasets including PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet. The proposed INSURE achieves state-of-the-art performance. We also empirically show that domain-specific class-relevant features are beneficial for domain generalization. The code is available at https://github.com/yuxi120407/INSURE.

Grid emergency voltage control (GEVC) is paramount in electric power systems to improve voltage stability and prevent cascading outages and blackouts in case of contingencies. While most deep reinforcement learning (DRL)-based paradigms perform single agents in a static environment, real-world agents for GEVC are expected to cooperate in a dynamically shifting grid. Moreover, due to high uncertainties from combinatory natures of various contingencies and load consumption, along with the complexity of dynamic grid operation, the data efficiency and control performance of the existing DRL-based methods are challenged. To address these limitations, we propose a multi-agent graph-attention (GATT)-based DRL algorithm for GEVC in multi-area power systems. We develop graph convolutional network (GCN)-based agents for feature representation of the graph-structured voltages to improve the decision accuracy in a data-efficient manner. Furthermore, a cutting-edge attention mechanism concentrates on effective information sharing among multiple agents, synergizing different-sized subnetworks in the grid for cooperative learning. We address several key challenges in the existing DRL-based GEVC approaches, including low scalability and poor stability against high uncertainties. Test results in the IEEE benchmark system verify the advantages of the proposed method over several recent multi-agent DRL-based algorithms.

Numerical simulation of non-linear partial differential equations plays a crucial role in modeling physical science and engineering phenomena, such as weather, climate, and aerodynamics. Recent Machine Learning (ML) models trained on low-resolution spatio-temporal signals have shown new promises in capturing important dynamics in high-resolution signals, under the condition that the models can effectively recover the missing details. However, this study shows that significant information is often lost in the low-resolution down-sampled features. To address such issues, we propose a new approach, namely Temporal Stencil Modeling (TSM), which combines the strengths of advanced time-series sequence modeling (with the HiPPO features) and state-of-the-art neural PDE solvers (with learnable stencil modeling). TSM aims to recover the lost information from the PDE trajectories and can be regarded as a temporal generalization of classic finite volume methods such as WENO. Our experimental results show that TSM achieves the new state-of-the-art simulation accuracy for 2-D incompressible Navier-Stokes turbulent flows: it significantly outperforms the previously reported best results by 19.9% in terms of the highly-correlated duration time, and reduces the inference latency into 80%. We also show a strong generalization ability of the proposed method to various out-of-distribution turbulent flow settings, as well as lower resolution or 1-D / 3-D settings. Our code is available at https://github.com/Edward-Sun/TSM-PDE .

Fast screening of drug molecules based on the ligand binding affinity is an important step in the drug discovery pipeline. Graph neural fingerprint is a promising method for developing molecular docking surrogates with high throughput and great fidelity. In this study, we built a COVID-19 drug docking dataset of about 300,000 drug candidates on 23 coronavirus protein targets. With this dataset, we trained graph neural fin-gerprint docking models for high-throughput virtual COVID-19 drug screening. The graph neural fingerprint models yield high prediction accuracy on docking scores with the mean squared error lower than 0.21 kcal/mol for most of the docking targets, showing significant improvement over conventional circular fin-gerprint methods. To make the neural fingerprints transferable for unknown targets, we also propose a transferable graph neural fingerprint method trained on multiple targets. With comparable accuracy to target-specific graph neural fingerprint models, the training and data efficiency of the transferable model is several times higher. We highlight that the impact of this study extends beyond COVID-19 dataset, as our approach for fast virtual ligand screening can be easily adapted and integrated into a general machine learning-accelerated pipeline to battle future bio-threats.

Modeling spatiotemporal brain dynamics from high-dimensional data, such as functional Magnetic Resonance Imaging (fMRI), is a formidable task in neuroscience. Existing approaches for fMRI analysis utilize hand-crafted features, but the process of feature extraction risks losing essential information in fMRI scans. To address this challenge, we present SwiFT (Swin 4D fMRI Transformer), a Swin Transformer architecture that can learn brain dynamics directly from fMRI volumes in a memory and computation-efficient manner. SwiFT achieves this by implementing a 4D window multi-head self-attention mechanism and absolute positional embeddings. We evaluate SwiFT using multiple large-scale resting-state fMRI datasets, including the Human Connectome Project (HCP), Adolescent Brain Cognitive Development (ABCD), and UK Biobank (UKB) datasets, to predict sex, age, and cognitive intelligence. Our experimental outcomes reveal that SwiFT consistently outperforms recent state-of-the-art models. Furthermore, by leveraging its end-to-end learning capability, we show that contrastive loss-based self-supervised pre-training of SwiFT can enhance performance on downstream tasks. Additionally, we employ an explainable AI method to identify the brain regions associated with sex classification. To our knowledge, SwiFT is the first Swin Transformer architecture to process dimensional spatiotemporal brain functional data in an end-to-end fashion. Our work holds substantial potential in facilitating scalable learning of functional brain imaging in neuroscience research by reducing the hurdles associated with applying Transformer models to high-dimensional fMRI.

Unpaired image-to-image translation has broad applications in art, design, and scientific simulations. One early breakthrough was CycleGAN that emphasizes one-to-one mappings between two unpaired image domains via generative-adversarial networks (GAN) coupled with the cycle-consistency constraint, while more recent works promote one-to-many mapping to boost diversity of the translated images. Motivated by scientific simulation and one-to-one needs, this work revisits the classic CycleGAN framework and boosts its performance to outperform more contemporary models without relaxing the cycle-consistency constraint. To achieve this, we equip the generator with a Vision Transformer (ViT) and employ necessary training and regularization techniques. Compared to previous best-performing models, our model performs better and retains a strong correlation between the original and translated image. An accompanying ablation study shows that both the gradient penalty and self-supervised pre-training are crucial to the improvement. To promote reproducibility and open science, the source code, hyperparameter configurations, and pre-trained model are available at https://github.com/LS4GAN/uvcgan.

Interpretation of cryo-electron microscopy (cryo-EM) maps requires building and fitting 3D atomic models of biological molecules. AlphaFold-predicted models generate initial 3D coordinates; however, model inaccuracy and conformational heterogeneity often necessitate labor-intensive manual model building and fitting into cryo-EM maps. In this work, we designed a protein model-building workflow, which combines a deep-learning cryo-EM map feature enhancement tool, CryoFEM (Cryo-EM Feature Enhancement Model) and AlphaFold. A benchmark test using 36 cryo-EM maps shows that CryoFEM achieves state-of-the-art performance in optimizing the Fourier Shell Correlations between the maps and the ground truth models. Furthermore, in a subset of 17 datasets where the initial AlphaFold predictions are less accurate, the workflow significantly improves their model accuracy. Our work demonstrates that the integration of modern deep learning image enhancement and AlphaFold may lead to automated model building and fitting for the atomistic interpretation of cryo-EM maps.

There are various sources of ionizing radiation exposure, where medical exposure for radiation therapy or diagnosis is the most common human-made source. Understanding how gene expression is modulated after ionizing radiation exposure and investigating the presence of any dose-dependent gene expression patterns have broad implications for health risks from radiotherapy, medical radiation diagnostic procedures, as well as other environmental exposure. In this paper, we perform a comprehensive pathway-based analysis of gene expression profiles in response to low-dose radiation exposure, in order to examine the potential mechanism of gene regulation underlying such responses. To accomplish this goal, we employ a statistical framework to determine whether a specific group of genes belonging to a known pathway display coordinated expression patterns that are modulated in a manner consistent with the radiation level. Findings in our study suggest that there exist complex yet consistent signatures that reflect the molecular response to radiation exposure, which differ between low-dose and high-dose radiation.

Today, voluminous multivariate time series data collected from sensors provides tremendous benefit for understanding of modern industrial systems such as power plants, wind turbines and aircrafts. However, the dynamic and complex nature of these systems, as well as the lack of prior knowledge impose challenges in perceiving different system behaviors from the time series data. To handle these issues, time series clustering has become one of the key analysis techniques. Nevertheless, the data nonlinearity, varying lengths and high dimensions of industrial time series could hinder the quality of clustering. To deal with these challenges, we propose Deep Time Series Sketching (DTSS) model. This model is a representation learning model based on temporal convolutional networks that perform on a sliding window basis along time series to learn the windows’ embeddings. The sequence of embeddings is then fed into embedding sketching to obtain its sketch. Such sketch is a descriptor of the whole time series and will be fed into K-means for clustering. Our model is a novel end-to-end hybrid model that incorporates both local and global contextual features. It is able to project multivariate time series with varying lengths into the same latent space. Moreover, we show that our model is able to perform early clustering as it can assign real-time label without seeing the whole time series. We test our model on both benchmark and real world industrial datasets, and experiments show that our proposed method outperforms popular time series clustering baselines.

Because protein-protein interactions (PPIs) are crucial to understand living systems, harvesting these data is essential to probe disease development and discern gene/protein functions and biological processes. Some curated datasets contain PPI data derived from the literature and other sources (e.g., IntAct, BioGrid, DIP, and HPRD). However, they are far from exhaustive, and their maintenance is a labor-intensive process. On the other hand, machine learning methods to automate PPI knowledge extraction from the scientific literature have been limited by a shortage of appropriate annotated data. This work presents a unified, multi-source PPI corpora with vetted interaction definitions augmented by binary interaction type labels and a Transformer-based deep learning method that exploits entities’ relational context information for relation representation to improve relation classification performance. The model’s performance is evaluated on four widely studied biomedical relation extraction datasets, as well as this work’s target PPI datasets, to observe the effectiveness of the representation to relation extraction tasks in various data. Results show the model outperforms prior state-of-the-art models. The code and data are available at: https://github.com/BNLNLP/PPI-Relation-Extraction

Long short-term memory (LSTM) is a kind of recurrent neural networks (RNN) for sequence and temporal dependency data modeling and its effectiveness has been extensively established. In this work, we propose a hybrid quantum-classical model of LSTM, which we dub QLSTM. We demonstrate that the proposed model successfully learns several kinds of temporal data. In particular, we show that for certain testing cases, this quantum version of LSTM converges faster, or equivalently, reaches a better accuracy, than its classical counterpart. Due to the variational nature of our approach, the requirements on qubit counts and circuit depth are eased, and our work thus paves the way toward implementing machine learning algorithms for sequence modeling such as natural language processing, speech recognition on noisy intermediate-scale quantum (NISQ) devices.

While deep neural networks have shown impressive performance in many tasks, they are fragile to carefully de-signed adversarial attacks. We propose a novel adversarial training-based model by Attention Guided Knowledge Distillation and Bi-directional Metric Learning (AGKD-BML). The attention knowledge is obtained from a weight-fixed model trained on a clean dataset, referred to as a teacher model, and transferred to a model that is under training on adversarial examples (AEs), referred to as a student model. In this way, the student model is able to focus on the correct region, as well as correcting the intermediate features corrupted by AEs to eventually improve the model accuracy. Moreover, to efficiently regularize the representation in feature space, we propose a bidirectional metric learning. Specifically, given a clean image, it is first attacked to its most confusing class to get the forward AE. A clean image in the most confusing class is then randomly picked and attacked back to the original class to get the backward AE. A triplet loss is then used to shorten the representation distance between original image and its AE, while enlarge that between the forward and backward AEs. We conduct extensive adversarial robustness experiments on two widely used datasets with different attacks. Our proposed AGKD-BML model consistently outperforms the state-of-the-art approaches. The code of AGKD-BML will be available at: https://github.com/hongw579/AGKD-BML.

Real-time data collection and analysis in large experimental facilities present a great challenge across multiple domains, including high energy physics, nuclear physics, and cosmology. To address this, machine learning (ML)-based methods for real-time data compression have drawn significant attention. However, unlike natural image data, such as CIFAR and ImageNet that are relatively small-sized and continuous, scientific data often come in as three-dimensional (3D) data volumes at high rates with high sparsity (many zeros) and non-Gaussian value distribution. This makes direct application of popular ML compression methods, as well as conventional data compression methods, suboptimal. To address these obstacles, this work introduces a dual-head autoencoder to resolve sparsity and regression simultaneously, called Bicephalous Convolutional AutoEncoder (BCAE). This method shows advantages both in compression fidelity and ratio compared to traditional data compression methods, such as MGARD, SZ, and ZFP. To achieve similar fidelity, the best performer among the traditional methods can reach only half the compression ratio of BCAE. Moreover, a thorough ablation study of the BCAE method shows that a dedicated segmentation decoder improves the reconstruction.

Clustering is a challenging problem when many features are irrelevant to separate clusters. Also, different clusters may relate to various feature subsets. This work proposes a deep clustering algorithm that localizes the search for instance clusters and their relevant features. The relevant features of each cluster are defined as those with high associations (dependency) within that cluster. Given the number of clusters$K$, we formulate the problem as$K$. -parallel auto-reconstructive learning, where low-rank graph learning, rooted in graph Laplacian theory, is used to explore the unknown feature associations of each cluster. The model performs automatic feature weighting on residuals to minimize loss from the corresponding cluster. Through such design, different feature subsets can be learned to calculate the loss from different clusters. Subsequently, the associations between features and clusters can be acquired, and better clustering result can be achieved. Moreover, the associated features of each cluster can be used to interpret the clustering patterns.

In today's industrial and scientific arenas, large quantities of multivariate time series data are generated without labels. Clustering such data is an important but challenging task due to complex variable associations. Unlike other previous efforts, this work explicitly explores variable associations through learning variable association graphs for each cluster. This is achieved through time series autoregression by a multi-path neural network, where each path corresponds to one cluster. The learned variable association graphs can be used to interpret how one cluster differs from another. Experiments demonstrate our framework's effectiveness on clustering and result interpretability.

Clustering is the task of instance grouping so that similar ones are grouped into the same cluster, while dissimilar ones are in different clusters. However, such similarity is a local concept in regard to different clusters and their relevant feature space. This work aims to discover clusters by exploring feature association and instance similarity concurrently. We propose a deep clustering framework that can localize the search for relevant features appertaining to different clusters. In turn, this allows for measuring instance similarity that exist in multiple, possibly overlapping, feature subsets, which contribute to more accurate clustering of instances. Additionally, the relevant features of each cluster endow interpretability of clustering results. Experiments on text and image datasets show that our method outperforms existing state-of-the-art baselines.

Identifying the faulty class of multivariate time series is crucial for today's flight data analysis. However, most of the existing time series classification methods suffer from imbalanced data and lack of model interpretability, especially on flight data of which faulty events are usually uncommon with a limited amount of data. Here, we present a neural network classification model for imbalanced multivariate time series by leveraging the information learned from normal class, which can also learn the nonlinear Granger causality for each class, so that we can pinpoint how time series classes differ from each other. Experiments on simulated data and real flight data shows that this model can achieve high accuracy of identifying anomalous flights.

Graph convolution is a generalization of the convolution operation from structured grid data to unstructured graph data. Because any type of data can be represented on a feature graph, graph convolution has been a powerful tool for modeling various types of data. However, such flexibility comes with a price: expensive time and space complexities. Even with state-of-the-art scalable graph convolution algorithms, it remains challenging to scale graph convolution for practical applications. Hence, we propose using Diverse Power Iteration Embeddings (DPIE) to construct scalable graph convolution neural networks. DPIE is an approximated spectral embedding with orders of magnitude faster speed that does not incur additional space complexity, resulting in efficient and effective graph convolution approximation. DPIE-based graph convolution avoids expensive convolution operation in the form of matrix-vector multiplication using the embedding of a lower dimension. At the same time, DPIE generates graphs implicitly, which dramatically reduces space cost when building graphs from unstructured data. The method is tested on various types of data. We also extend the graph convolution to extreme-scale data never-before studied in the graph convolution field. Experiment results show the scalability and effectiveness of DPIE-based graph convolution.

This work examines the convergence of stochastic gradient-based optimization algorithms that use early stopping based on a validation function. The form of early stopping we consider is that optimization terminates when the norm of the gradient of a validation function falls below a threshold. We derive conditions that guarantee this stopping rule is well-defined, and provide bounds on the expected number of iterations and gradient evaluations needed to meet this criterion. The guarantee accounts for the distance between the training and validation sets, measured with the Wasserstein distance. We develop the approach in the general setting of a first-order optimization algorithm, with possibly biased update directions subject to a geometric drift condition. We then derive bounds on the expected running time for early stopping variants of several algorithms, including stochastic gradient descent (SGD), decentralized SGD (DSGD), and the stochastic variance reduced gradient (SVRG) algorithm. Finally, we consider the generalization properties of the iterate returned by early stopping.

System deterioration detection and root cause analysis is crucial for today's industrial society. However, the design and operation of mechanic system is getting more and more complex, which makes it hard at identifying deterioration with noisy data. Our research focuses on solving such problem on time-evolving sensor graphs in a streaming setting. Given a sequence of graphs, the ability to identify 1) any gradual and stable structured change and 2) the root cause components is of importance for early warning and system diagnosis. Existing methods either raise too many false alerts on instant changes or are too sensitive to noise. To address these problems, we propose Robust Failure Detection and Diagnosis (RoFaD). RoFaD can capture failure propagation given a time series of graph. By optimizing a matrix-based Taylor expansion, RoFaD can identify system deterioration in the presence of noise and immediate changes, and diagnose the root cause components. Experiments on both synthetic and real world datasets demonstrate that RoFaD is more effective than the popular baselines.

Learning the causal graph in a complex system is crucial for knowledge discovery and decision making, yet it remains a challenging problem because of the unknown nonlinear interaction among system components. Most of the existing methods either rely on predefined kernel or data distribution, or they focus simply on the causality between a single target and the remaining system. This work presents a deep neural network for scalable causal graph learning (SCGL) through low-rank approximation. The SCGL model can explore nonlinearity on both temporal and intervariable relationships without any predefined kernel or distribution assumptions. Through low-rank approximation, the noise influence is reduced, and better accuracy and high scalability are achieved. Experiments using synthetic and real-world datasets show that our SCGL algorithm outperforms existing state-of-the-art methods for causal graph learning.

Carlos Soto, Shinjae Yoo. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Discovering the causal graph in multivariate time series data is of great importance for industrial society, yet challenging due to the unknown nonlinearity in the data. Existing works only explore the data in chronological order, and rely on pre-assumed kernels or certain distribution assumption. In this paper, we present a Bi-directional neural network for Causal Graph Learning (Bi-CGL) through weight-sharing and low-rank neural network. It discovers the causal graph by simultaneously exploring input in forward and reverse chronological order. Both directions approach the same causal graph with shared low-rank approximation, which provides robustness and better accuracy against data noise. Experiments on synthetic and real world datasets prove our Bi-CGL's outperformance over existing baselines.

With latent variables, stochastic recurrent models have achieved state-of-the-art performance in modeling sound-wave sequence. However, opposite results are also observed in other domains, where standard recurrent networks often outperform stochastic models. To better understand this discrepancy, we re-examine the roles of latent variables in stochastic recurrent models for speech density estimation. Our analysis reveals that under the restriction of fully factorized output distribution in previous evaluations, the stochastic variants were implicitly leveraging intra-step correlation but the deterministic recurrent baselines were prohibited to do so, resulting in an unfair comparison. To correct the unfairness, we remove such restriction in our re-examination, where all the models can explicitly leverage intra-step correlation with an auto-regressive structure. Over a diverse set of univariate and multivariate sequential data, including human speech, MIDI music, handwriting trajectory, and frame-permuted speech, our results show that stochastic recurrent models fail to deliver the performance advantage claimed in previous work. %exhibit any practical advantage despite the claimed theoretical superiority. In contrast, standard recurrent models equipped with an auto-regressive output distribution consistently perform better, dramatically advancing the state-of-the-art results on three speech datasets.

In this era of Big Data, computing useful and timely information from data is becoming increasingly complicated, particularly due to the ever increasing volumes of data that need to travel over the network to data centers to be stored and processed, all highly expensive operations in the long haul. This is a strong motivation to explore how to perform computing and analysis of data “on the wire”, i.e., while the data is still in transit. The nature of these computations include analysis, visualization, pattern recognition, and prediction on the streaming data. In this paper we present the idea of a framework capable of analyzing data in transit based on the principles of a service function chaining architecture. This framework can be deployed at any practical location within the network where computation on data flows is desirable. We further describe an all-virtual implementation of the framework as a worst-case scenario and present an early investigation of its capabilities with three examples — pattern recognition and data visualization on streaming Forex data, targeted advertising from clickstream data, and processing of monitoring data from solar sensors for maintenance decisions. Our results indicate that performing computations on live data flows to provide immediate perspective on the data is possible and attractive, but also that performance heavily depends on the amount and capabilities of the dedicated resources.

Clustering is a classical data mining task used for discovering interrelated pattern of similarities in the data. In many modern day domains, data is getting continuously generated as a stream. For scalability reasons, clustering the points in a data stream requires designing single pass, limited memory streaming clustering algorithms. However, the performance of the known streaming clustering algorithms, that typically use K-means (or its variants) on the original feature space, tend to suffer when the feature space is high-dimensional. To overcome this problem, we propose a streaming spectral clustering algorithm. Our algorithm maintains an approximation of the normalized Laplacian of the data stream over time and efficiently updates the changing eigenvectors of this Laplacian in a streaming fashion. It requires just one pass over the data, consumes limited memory, and is stable to the ordering of the data stream. We provide a theoretical analysis of our streaming spectral clustering algorithm and our experimental results show that while gaining in scalability, its performance compares well with other known batch/streaming clustering approaches.

Manifold learning, especially spectral embedding, is known as one of the most effective learning approaches on high dimensional data, but for real-world applications it raises a serious computational burden in constructing spectral embeddings for large datasets. To overcome this computational complexity, we propose a novel efficient embedding construction, Diverse Power Iteration Embedding (DPIE). DPIE shows almost the same effectiveness of spectral embeddings and yet is three order of magnitude faster than spectral embeddings computed from eigen-decomposition. Our DPIE is unique in that (1) it finds linearly independent embeddings and thus shows diverse aspects of dataset; (2) the proposed regularized DPIE is effective if we need many embeddings; (3) we show how to efficiently orthogonalize DPIE if one needs; and (4) Diverse Power Iteration Value (DPIV) provides the importance of each DPIE like an eigen value. Such various aspects of DPIE and DPIV ensure that our algorithm is easy to apply to various applications, and we also show the effectiveness and efficiency of DPIE on clustering, anomaly detection, and feature selection as our case studies.

Massive data streams are continuously being generated from sources such as social media, broadcast news, etc., and typically these datapoints lie in high-dimensional spaces (such as the vocabulary space of a language). Timely and accurate feature subset selection in these massive data streams has important applications in model interpretation, computational/storage cost reduction, and generalization enhancement. In this paper, we introduce a novel unsupervised feature selection approach on data streams that selects important features by making only one pass over the data while utilizing limited storage. The proposed algorithm uses ideas from matrix sketching to efficiently maintain a low-rank approximation of the observed data and applies regularized regression on this approximation to identify the important features. We theoretically prove that our algorithm is close to an expensive offline approach based on global singular value decompositions. The experimental results on a variety of text and image datasets demonstrate the excellent ability of our approach to identify important features even in presence of concept drifts and also its efficiency over other popular scalable feature selection algorithms.

BACKGROUND: Our study focuses on discovering gene regulatory networks from time series gene expression data using the Granger causality (GC) model. However, the number of available time points (T) usually is much smaller than the number of target genes (n) in biological datasets. The widely applied pairwise GC model (PGC) and other regularization strategies can lead to a significant number of false identifications when n>>T. RESULTS: In this study, we proposed a new method, viz., CGC-2SPR (CGC using two-step prior Ridge regularization) to resolve the problem by incorporating prior biological knowledge about a target gene data set. In our simulation experiments, the propose new methodology CGC-2SPR showed significant performance improvement in terms of accuracy over other widely used GC modeling (PGC, Ridge and Lasso) and MI-based (MRNET and ARACNE) methods. In addition, we applied CGC-2SPR to a real biological dataset, i.e., the yeast metabolic cycle, and discovered more true positive edges with CGC-2SPR than with the other existing methods. CONCLUSIONS: In our research, we noticed a " 1+1>2" effect when we combined prior knowledge and gene expression data to discover regulatory networks. Based on causality networks, we made a functional prediction that the Abm1 gene (its functions previously were unknown) might be related to the yeast's responses to different levels of glucose. Our research improves causality modeling by combining heterogeneous knowledge, which is well aligned with the future direction in system biology. Furthermore, we proposed a method of Monte Carlo significance estimation (MCSE) to calculate the edge significances which provide statistical meanings to the discovered causality networks. All of our data and source codes will be available under the link https://bitbucket.org/dtyu/granger-causality/wiki/Home.

Current spectral clustering algorithms suffer from the sensitivity to existing noise and parameter scaling and may not be aware of different density distributions across clusters. If these problems are left untreated, the consequent clustering results cannot accurately represent true data patterns, in particular, for complex real-world datasets with heterogeneous densities. This article aims to solve these problems by proposing a diffusion-based Aggregated Heat Kernel (AHK) to improve the clustering stability, and a Local Density Affinity Transformation (LDAT) to correct the bias originating from different cluster densities. AHK statistically models the heat diffusion traces along the entire time scale, so it ensures robustness during the clustering process, while LDAT probabilistically reveals the local density of each instance and suppresses the local density bias in the affinity matrix. Our proposed framework integrates these two techniques systematically. As a result, it not only provides an advanced noise-resisting and density-aware spectral mapping to the original dataset but also demonstrates the stability during the processing of tuning the scaling parameter (which usually controls the range of neighborhood). Furthermore, our framework works well with the majority of similarity kernels, which ensures its applicability to many types of data and problem domains. The systematic experiments on different applications show that our proposed algorithm outperforms state-of-the-art clustering algorithms for the data with heterogeneous density distributions and achieves robust clustering performance with respect to tuning the scaling parameter and handling various levels and types of noise.

Spectral Embedding is one of the most effective dimension reduction algorithms in data mining. However, its computation complexity has to be mitigated in order to apply it for real-world large scale data analysis. Many researches have been focusing on developing approximate spectral embeddings which are more efficient, but meanwhile far less effective. This paper proposes Diverse Power Iteration Embeddings (DPIE), which not only retains the similar efficiency of power iteration methods but also produces a series of diverse and more effective embedding vectors. We test this novel method by applying it to various data mining applications (e.g. Clustering, anomaly detection and feature selection) and evaluating their performance improvements. The experimental results show our proposed DPIE is more effective than popular spectral approximation methods, and obtains the similar quality of classic spectral embedding derived from eigen-decompositions. Moreover it is extremely fast on big data applications. For example in terms of clustering result, DPIE achieves as good as 95% of classic spectral clustering on the complex datasets but 4000+ times faster in limited memory environment.

Unsupervised feature selection is an important issue for high dimensional dataset analysis. However popular methods are susceptible to noisy instances (observations) or noisy features. We propose a noise-resistant feature selection algorithm by capturing multi-perspective correlations. Our proposed approach, called Noise-Resistant Unsupervised Feature Selection (NRFS), is based on multi-perspective correlation that reflects the importance of feature with respect to noise-resistant representative instances and various global trends from spectral decomposition. In this way, the model concisely captures a wide variety of local patterns. Experimental results demonstrate the effectiveness of our algorithm.

Current popular anomaly detection algorithms are capable of detecting global anomalies but often fail to distinguish local anomalies from normal instances. Inspired by contemporary physics theory (i.e., heat diffusion and quantum mechanics), we propose two unsupervised anomaly detection algorithms. Building on the embedding manifold derived from heat diffusion, we devise Local Anomaly Descriptor (LAD), which faithfully reveals the intrinsic neighborhood density. It uses a scale-dependent umbrella operator to bridge global and local properties, which makes LAD more informative within an adaptive scope of neighborhood. To offer more stability of local density measurement on scaling parameter tuning, we formulate Fermi Density Descriptor (FDD), which measures the probability of a fermion particle being at a specific location. By choosing the stable energy distribution function, FDD steadily distinguishes anomalies from normal instances with any scaling parameter setting. To further enhance the efficacy of our proposed algorithms, we explore the utility of anisotropic Gaussian kernel (AGK), which offers better manifold-aware affinity information. We also quantify and examine the effect of different Laplacian normalizations for anomaly detection. Comprehensive experiments on both synthetic and benchmark datasets verify that our proposed algorithms outperform the existing anomaly detection algorithms.

Current popular anomaly detection algorithms are capable of detecting global anomalies but oftentimes fail to distinguish local anomalies from normal instances. This paper aims to improve unsupervised anomaly detection via the exploration of physics-based diffusion space. Building upon the embedding manifold derived from diffusion maps, we devise Local Anomaly Descriptor (LAD) whose originality results from faithfully preserving intrinsic and informative density-relevant neighborhood information. This robust and effective algorithm is designed with a weighted umbrella Laplacian operator to bridge global and local properties. To further enhance the efficacy of our proposed algorithm, we explore the utility of anisotropic Gaussian kernel (AGK) which can offer better manifold-aware affinity information. Comprehensive experiments on both synthetic and UCI real datasets verify that our LAD outperforms existing anomaly detection algorithms.

The primary originality of this paper lies at the fact that we have made the first attempt to apply quantum mechanics theory to anomaly (outlier) detection in high-dimensional datasets for data mining. We propose Fermi Density Descriptor (FDD) which represents the probability of measuring a fermion at a specific location for anomaly detection. We also quantify and examine different Laplacian normalization effects and choose the best one for anomaly detection. Both theoretical proof and quantitative experiments demonstrate that our proposed FDD is substantially more discriminative and robust than the commonly-used algorithms.

Email overload, even after spam filtering, presents a serious productivity challenge for busy professionals and executives. One solution is automated prioritization of incoming emails to ensure the most important are read and processed quickly, while others are processed later as/if time permits in declining priority levels. This paper presents a study of machine learning approaches to email prioritization into discrete levels, comparing ordinal regression versus classifier cascades. Given the ordinal nature of discrete email priority levels, SVM ordinal regression would be expected to perform well, but surprisingly a cascade of SVM classifiers significantly outperforms ordinal regression for email prioritization. In contrast, SVM regression performs well -- better than classifiers -- on selected UCI data sets. This unexpected performance inversion is analyzed and results are presented, providing core functionality for email prioritization systems.

Current spectral clustering algorithms suffer from both sensitivity to scaling parameter selection in similarity matrix construction, and data perturbation. This paper aims to improve robustness in clustering algorithms and combat these two limitations based on heat kernel theory. Heat kernel can statistically depict traces of random walk, so it has an intrinsic connection with diffusion distance, with which we can ensure robustness during any clustering process. By integrating heat distributed along time scale, we propose a novel method called Aggregated Heat Kernel (AHK) to measure the distance between each point pair in their eigen space. Using AHK and Laplace-Beltrami Normalization (LBN) we are able to apply an advanced noise-resisting robust spectral mapping to original dataset. Moreover it offers stability on scaling parameter tuning. Experimental results show that, compared to other popular spectral clustering methods, our algorithm can achieve robust clustering results on both synthetic and UCI real datasets.

Email is one of the most prevalent communication tools today, and solving the email overload problem is pressingly urgent. A good way to alleviate email overload is to automatically prioritize received messages according to the priorities of each user. However, research on statistical learning methods for fully personalized email prioritization (PEP) has been sparse due to privacy issues, since people are reluctant to share personal messages and importance judgments with the research community. It is therefore important to develop and evaluate PEP methods under the assumption that only limited training examples can be available, and that the system can only have the personal email data of each user during the training and testing of the model for that user. This paper presents the first study (to the best of our knowledge) under such an assumption. Specifically, we focus on analysis of personal social networks to capture user groups and to obtain rich features that represent the social roles from the viewpoint of a particular user. We also developed a novel semi-supervised (transductive) learning algorithm that propagates importance labels from training examples to test examples through message and user nodes in a personal email network. These methods together enable us to obtain an enriched vector representation of each new email message, which consists of both standard features of an email message (such as words in the title or body, sender and receiver IDs, etc.) and the induced social features from the sender and receivers of the message. Using the enriched vector representation as the input in SVM classifiers to predict the importance level for each test message, we obtained significant performance improvement over the baseline system (without induced social features) in our experiments on a multi-user data collection. We obtained significant performance improvement over the baseline system (without induced social features) in our experiments on a multi-user data collection: the relative error reduction in MAE was 31% in micro-averaging, and 14% in macro-averaging.

This paper reports a cross-benchmark evaluation of regularized logistic regression (LR) and incremental Rocchio for adaptive filtering. Using four corpora from the Topic Detection and Tracking (TDT) forum and the Text Retrieval Conferences (TREC) we evaluated these methods with non-stationary topics at various granularity levels, and measured performance with different utility settings. We found that LR performs strongly and robustly in optimizing T11SU (a TREC utility function) while Rocchio is better for optimizing Ctrk (the TDT tracking cost), a high-recall oriented objective function. Using systematic cross-corpus parameter optimization with both methods, we obtained the best results ever reported on TDT5, TREC10 and TREC11. Relevance feedback on a small portion (0.05~0.2%) of the TDT5 test documents yielded significant performance improvements, measuring up to a 54% reduction in Ctrk and a 20.9% increase in T11SU (with b=0.1), compared to the results of the top-performing system in TDT2004 without relevance feedback information.