Feature Transformation is crucial for classic machine learning that aims to generate feature combinations to enhance the performance of downstream tasks from a data-centric perspective. Current methodologies, such as manual expert-driven processes, iterative-feedback techniques, and exploration-generative tactics, have shown promise in automating such data engineering workflow by minimizing human involvement. However, three challenges remain in those frameworks: (1) It predominantly depends on downstream task performance metrics, as assessment is time-consuming, especially for large datasets. (2) The diversity of feature combinations will hardly be guaranteed after random exploration ends. (3) Rare significant transformations lead to sparse valuable feedback that hinders the learning processes or leads to less effective results. In response to these challenges, we introduce FASTFT, an innovative framework that leverages a trio of advanced strategies. We first decouple the feature transformation evaluation from the outcomes of the generated datasets via the performance predictor. To address the issue of reward sparsity, we developed a method to evaluate the novelty of generated transformation sequences. Incorporating this novelty into the reward function accelerates the model's exploration of effective transformations, thereby improving the search productivity. Additionally, we combine novelty and performance to create a prioritized memory buffer, ensuring that essential experiences are effectively revisited during exploration. Our extensive experimental evaluations validate the performance, efficiency, and traceability of our proposed framework, showcasing its superiority in handling complex feature transformation tasks11The code and data are publicly accessible via Github..

In recent years, the rapid advancement of Artificial Intelligence (AI) technologies, particularly Large Language Models (LLMs), has revolutionized the paradigm of scientific discovery, establishing AI-for-Science (AI4Science) as a dynamic and evolving field. However, there is still a lack of an effective framework for the overall assessment of AI4Science, particularly from a holistic perspective on data quality and model capability. Therefore, in this study, we propose SciHorizon, a comprehensive assessment framework designed to benchmark the readiness of AI4Science from both scientific data and LLM perspectives. First, we introduce a generalizable framework for assessing AI-ready scientific data, encompassing four key dimensions-Quality, FAIRness, Explainability, and Compliance-which are subdivided into 15 sub-dimensions. Drawing on data resource papers published between 2018 and 2023 in peer-reviewed journals, we present recommendation lists of AI-ready datasets for Earth, Life, and Materials Sciences, making a novel and original contribution to the field. Concurrently, to assess the capabilities of LLMs across multiple scientific disciplines, we establish 16 assessment dimensions based on five core indicators-Knowledge, Understanding, Reasoning, Multimodality, and Values-spanning Mathematics, Physics, Chemistry, Life Sciences, and Earth and Space Sciences. Using the developed benchmark datasets, we have conducted a comprehensive evaluation of over 50 representative open-source and closed-source LLMs. All the results are publicly available and can be accessed online at www.scihorizon.cn/en.

Zero-shot hashing (ZSH) has shown excellent success owing to its efficiency and generalization in large-scale retrieval scenarios. However, existing works ignore the locality relationships of representations and attributes, which have effective transferability between seeable classes and unseeable classes. Also, the continuous value attributes are not fully harnessed. In response, we conduct a COMprehensive Attribute Exploration for ZSH, named COMAE, which depicts the relationships from seen classes to unseen ones through three meticulously designed explorations, i.e., point-wise, pair-wise and class-wise consistency constraints. By regressing attributes from the proposed attribute prototype network, COMAE learns the local features that are relevant to the visual attributes. Then COMAE utilizes contrastive learning to comprehensively depict the context of attributes, rather than instance-independent optimization. Finally, the class-wise constraint is designed to cohesively learn the hash code, image representation, and visual attributes more effectively. Furthermore, theoretical analysis is provided to show the effectiveness of COMAE. Experimental results demonstrate that COMAE outperforms state-of-the-art hashing models, especially in scenarios with a larger number of unseen label classes.

Large language models (LLMs) have shown state-of-the-art results in translating natural language questions into SQL queries (Text-to-SQL), a long-standing challenge within the database community. However, security concerns remain largely unexplored, particularly the threat of backdoor attacks, which can introduce malicious behaviors into models through fine-tuning with poisoned datasets. In this work, we systematically investigate the vulnerabilities of LLM-based Text-to-SQL models and present ToxicSQL, a novel backdoor attack framework. Our approach leverages stealthy command-like and character-level triggers to make backdoors difficult to detect and remove, ensuring that malicious behaviors remain covert while maintaining high model accuracy on benign inputs. Furthermore, we propose leveraging SQL injection payloads as backdoor targets, enabling the generation of malicious yet executable SQL queries, which pose severe security and privacy risks in language model-based SQL development. We demonstrate that injecting only 0.44% of poisoned data can result in an attack success rate of 79.41%, posing a significant risk to database security. Additionally, we propose detection and mitigation strategies to enhance model reliability. Our findings highlight the urgent need for security-aware Text-to-SQL development, emphasizing the importance of robust defenses against backdoor threats.

Transformer-based large language models (LLMs) have demonstrated outstanding performance across diverse domains, particularly in the emerging pretrain-then-finetune paradigm. LoRA, a parameter-efficient fine-tuning method, is commonly used to adapt a base LLM to multiple downstream tasks. Further, LLM platforms enable developers to fine-tune multiple models and develop various domain-specific applications simultaneously. However, existing model parallelism schemes suffer from high communication overhead and inefficient GPU utilization. In this paper, we present mLoRA, a parallelism-efficient fine-tuning system designed for training multiple LoRA across GPUs and machines. mLoRA introduces a novel LoRA-aware pipeline parallelism scheme that efficiently pipelines LoRA adapters and their distinct fine-tuning stages across GPUs and machines, along with a new LoRA-efficient operator to enhance GPU utilization. Our extensive evaluation shows that mLoRA can significantly reduce average fine-tuning task completion time, e.g., by 30%, compared to state-of-the-art methods like FSDP. More importantly, mLoRA enables simultaneous fine-tuning of larger models, e.g., two Llama-2-13B models on four NVIDIA RTX A6000 48GB GPUs, which is not feasible for FSDP due to high memory requirements. Hence, mLoRA not only increases fine-tuning efficiency but also makes it more accessible on cost-effective GPUs.

Researching genes and their interactions is crucial for deciphering the fundamental laws of cellular activity, advancing disease treatment, drug discovery, and more. Large language Models (LLMs), with their profound text comprehension and generation capabilities, have made significant strides across various natural science fields. However, their application in cell biology remains limited and a systematic evaluation of their performance is lacking. To address this gap, in this article, we select seven mainstream LLMs and evaluate their performance across nine gene-related problem scenarios. Our findings indicate that LLMs possess a certain level of understanding of genes and cells, but still lag behind domain-specific models in comprehending transcriptional expression profiles. Moreover, we have improved the current method of textual representation of cells, enhancing the LLMs’ ability to tackle cell annotation tasks. We encourage cell biology researchers to leverage LLMs for problem-solving while being mindful of the associated challenges. We release our code and data at https://github.com/epang-ucas/Evaluate_LLMs_to_Genes .

The objective of topic inference in research proposals aims to obtain the most suitable disciplinary division from the discipline system defined by a funding agency. The agency will subsequently find appropriate peer-review experts from their database based on this division. Automated topic inference can reduce human errors caused by manual topic filling, bridge the knowledge gap between funding agencies and project applicants, and improve system efficiency. Existing methods focus on modeling this as a hierarchical multi-label classification problem, using generative models to iteratively infer the most appropriate topic information. However, these methods overlook the gap in scale between interdisciplinary research proposals and non-interdisciplinary ones, leading to an unjust phenomenon where the automated inference system categorizes interdisciplinary proposals as non-interdisciplinary, causing unfairness during the expert assignment. How can we address this data imbalance issue under a complex discipline system and hence resolve this unfairness? In this article, we implement a topic label inference system based on a Transformer encoder–decoder architecture. Furthermore, we utilize interpolation techniques to create a series of pseudo-interdisciplinary proposals from non-interdisciplinary ones during training based on non-parametric indicators, such as cross-topic probabilities and topic occurrence probabilities. This approach aims to reduce the bias of the system during model training. Finally, we conduct extensive experiments on a real-world dataset to verify the effectiveness of the proposed method. The experimental results demonstrate that our training strategy can significantly mitigate the unfairness generated in the topic inference task. To improve the reproducibility of our research, we have released accompanying code by Dropbox. 1

Accurately predicting Drug-Drug Interactions (DDIs) is critical to designing effective drug combination therapies. Recently, Artificial Intelligence (AI)-powered DDI prediction approaches have emerged as a new paradigm. However, most existing methods oversimplify the complex hierarchical structure within molecules and overlook the multi-source heterogeneous information external to molecules, limiting their modeling and predictive capabilities. To address this, we propose a H ierarchical H eterogeneous graph learning framework for D DI prediction, namely H2D. H2D employs an internal-to-external, local-to-global hierarchical perspective, exploiting intra-molecular multi-granularity structures and inter-molecular biomedical interactions to mutually enhance across hierarchical levels. Extensive experimental results demonstrate H2D's effectiveness on three real-world DDI prediction tasks (binary-class, multi-class, and multi-label). In sum, H2D achieves state-of-the-art performance in DDI prediction by leveraging the multi-scale graph structures, opening up new avenues in AI-powered DDI prediction.

Zero-Shot Hashing (ZSH) has aroused significant attention due to its efficiency and generalizability in multi-modal retrieval scenarios, which aims to encode semantic information into hash codes without needing unseen labeled training samples. In addition to commonly used visual images as visual semantics and class labels as global semantics, the corresponding attribute descriptions contain critical local semantics with detailed information. However, most existing methods focus on leveraging the extracted attribute numerical values, without exploring the textual semantics in attribute descriptions. To bridge this gap, in this paper, we propose Prompt-based zero-shot hashing via vIsual and teXtual sEmantic aLignment, namely PIXEL. Concretely, we design the attribute prompt template depending on attribute descriptions to make the model capture the corresponding local semantics. Then, achieving the textual embedding and visual embedding, we proposed an alignment module to model the intra- and inter-class contrastive distances. In addition, the attribute-wise constraint and class-wise constraint are utilized to collaboratively learn the hash code, image representation, and visual attributes more effectively. Finally, extensive experimental results demonstrate the superiority of PIXEL.

Multimodal recommendation systems (MMRS) have received considerable attention from the research community due to their ability to jointly utilize information from user behavior and product images and text. Previous research has two main issues. First, many long-tail items in recommendation systems have limited interaction data, making it difficult to learn comprehensive and informative representations. However, past MMRS studies have overlooked this issue. Secondly, users' modality preferences are crucial to their behavior. However, previous research has primarily focused on learning item modality representations, while user modality representations have remained relatively simplistic. To address these challenges, we propose a novel Graphs and User Modalities Enhancement (GUME) for long-tail multimodal recommendation. Specifically, we first enhance the user-item graph using multimodal similarity between items. This improves the connectivity of long-tail items and helps them learn high-quality representations through graph propagation. Then, we construct two types of user modalities: explicit interaction features and extended interest features. By using the user modality enhancement strategy to maximize mutual information between these two features, we improve the generalization ability of user modality representations. Additionally, we design an alignment strategy for modality data to remove noise from both internal and external perspectives. Extensive experiments on four publicly available datasets demonstrate the effectiveness of our approach. The code and data are publicly accessible via GitHub.

Molecular property prediction stands as a cornerstone task in AI-driven drug design and discovery, wherein the atoms within a molecule serve as nodes, collectively forming a graph with bonds acting as edges. Given the crucial role of geometric structures in molecular property prediction, the integration of 3D information with various graph learning methods has been explored to enhance prediction performance. Despite the increasing adoption of the "Graph pre-training and fine-tuning" paradigm to refine molecular representations, a significant challenge persists due to the misalignment between pre-training objectives and downstream tasks. Drawing inspiration from prompt tuning techniques in Natural Language Processing (NLP), several graph prompt-based methods have emerged. However, existing approaches tend to overlook the unique properties inherent in molecular graphs. To address this gap, our paper introduces a novel approach named 3D MO lecul A rpromp T (MOAT) designed specifically for geometric molecules. Specifically, we propose atom-level prompts to capture atom distribution, geometry-level prompts tailored for molecular conformers, where different conformations have distinct chemical properties, and task-level prompts to leverage functional group properties. Results on both 3D and 2D downstream tasks demonstrate its ability to successfully bridge the data gap across diverse settings. To the best of our knowledge, this paper is the first attempt to introduce geometric graph-prompting learning for molecules.

Feature transformation is to derive a new feature set from original features to augment the AI power of data. In many science domains such as material performance screening, while feature transformation can model material formula interactions and compositions and discover performance drivers, supervised labels are collected from expensive and lengthy experiments. This issue motivates an Unsupervised Feature Transformation Learning (UFTL) problem. Prior literature, such as manual transformation, supervised feedback guided search, and PCA, either relies on domain knowledge or expensive supervised feedback, or suffers from large search space, or overlooks non-linear feature-feature interactions. UFTL imposes a major challenge on existing methods: how to design a new unsupervised paradigm that captures complex feature interactions and avoids large search space? To fill this gap, we connect graph, contrastive, and generative learning to develop a measurement-pretrain-finetune paradigm for UFTL. For unsupervised feature set utility measurement, we propose a feature value consistency preservation perspective and develop a mean discounted cumulative gain like unsupervised metric to evaluate feature set utility. For unsupervised feature set representation pretraining, we regard a feature set as a feature-feature interaction graph, and develop an unsupervised graph contrastive learning encoder to embed feature sets into vectors. For generative transformation finetuning, we regard a feature set as a feature cross sequence and feature transformation as sequential generation. We develop a deep generative feature transformation model that coordinates the pretrained feature set encoder and the gradient information extracted from a feature set utility evaluator to optimize a transformed feature generator. Finally, we conduct extensive experiments to demonstrate the effectiveness, efficiency, traceability, and explicitness of our framework.

Traffic flow forecasting is a fundamental research issue for transportation planning and management, which serves as a canonical and typical example of spatial-temporal predictions. In recent years, Graph Neural Networks (GNNs) and Recurrent Neural Networks (RNNs) have achieved great success in capturing spatial-temporal correlations for traffic flow forecasting. Yet, two non-ignorable issues haven't been well solved: 1) The message passing in GNNs is immediate, while in reality the spatial message interactions among neighboring nodes can be delayed. The change of traffic flow at one node will take several minutes, i.e., time delay, to influence its connected neighbors. 2) Traffic conditions undergo continuous changes. The prediction frequency for traffic flow forecasting may vary based on specific scenario requirements. Most existing discretized models require retraining for each prediction horizon, restricting their applicability. To tackle the above issues, we propose a neural Spatial-Temporal Delay Differential Equation model, namely STDDE. It includes both delay effects and continuity into a unified delay differential equation framework, which explicitly models the time delay in spatial information propagation. Furthermore, theoretical proofs are provided to show its stability. Then we design a learnable traffic-graph time-delay estimator, which utilizes the continuity of the hidden states to achieve the gradient backward process. Finally, we propose a continuous output module, allowing us to accurately predict traffic flow at various frequencies, which provides more flexibility and adaptability to different scenarios. Extensive experiments show the superiority of STDDE. Both quantitative and qualitative experiments are conducted to validate the concept of a delay-aware module. Also, the flexibility validation shows the effectiveness of the continuous output module.

Deep learning recommendation models (DLRM) rely on large embedding tables to manage categorical sparse features. Expanding such embedding tables can significantly enhance model performance, but at the cost of increased GPU/CPU/memory usage. Meanwhile, tech companies have built extensive cloud-based services to accelerate training DLRM models at scale. In this paper, we conduct a deep investigation of the DLRM training platforms at AntGroup and reveal two critical challenges: low resource utilization due to suboptimal configurations by users and the tendency to encounter abnormalities due to an unstable cloud environment. To overcome them, we introduce DLRover, an elastic training framework for DLRMs designed to increase resource utilization and handle the instability of a cloud environment. DLRover develops a resource-performance model by considering the unique characteristics of DLRMs and a three-stage heuristic strategy to automatically allocate and dynamically adjust resources for DLRM training jobs for higher resource utilization. Further, DLRover develops multiple mechanisms to ensure efficient and reliable execution of DLRM training jobs. Our extensive evaluation shows that DLRover reduces job completion times by 31%, increases the job completion rate by 6%, enhances CPU usage by 15%, and improves memory utilization by 20%, compared to state-of-the-art resource scheduling frameworks. DLRover has been widely deployed at AntGroup and processes thousands of DLRM training jobs on a daily basis. DLRover is open-sourced and has been adopted by 10+ companies.

Wind is one kind of high-efficient, environmentally-friendly, and cost-effective energy source. Wind power, as one of the largest renewable energy in the world, has been playing a more and more important role in supplying electricity. Though growing dramatically in recent years, the amount of generated wind power can be directly or latently affected by multiple uncertain factors, such as wind speed, wind direction, temperatures, and so on. More importantly, there exist very complicated dependencies of the generated power on the latent composition of these multiple time-evolving variables, which are always ignored by existing works and thus largely hinder the prediction performances. To this end, we propose DEWP , a novel D eep E xpansion learning for W ind P ower forecasting framework to carefully model the complicated dependencies with adequate expressiveness. DEWP starts with a stack-by-stack architecture, where each stack is composed of (i) a variable expansion block that makes use of convolutional layers to capture dependencies among multiple variables; (ii) a time expansion block that applies Fourier series and backcast/forecast mechanism to learn temporal dependencies in sequential patterns. These two tailored blocks expand raw inputs into different latent feature spaces which can model different levels of dependencies of time-evolving sequential data. Moreover, we propose an inference block corresponding for each stack, which applies multi-head self-attentions to acquire attentive features and maps expanded latent representations into generated wind power. In addition, to make DEWP more expressive in handling deep neural architectures, we adapt doubly residue learning to process stack-by-stack outputs. Accurate wind power forecasting (WPF) is then better achieved through fine-grained outputs by continuously removing stack residues and accumulating useful stack forecasts. Finally, we present extensive experiments in the real-world WPF application on two datasets from two different turbines, in order to demonstrate the effectiveness of our approach.

Feature transformation aims to reconstruct an effective representation space by mathematically refining the existing features. It serves as a pivotal approach to combat the curse of dimensionality, enhance model generalization, mitigate data sparsity, and extend the applicability of classical models. Existing research predominantly focuses on domain knowledge-based feature engineering or learning latent representations. However, these methods, while insightful, lack full automation and fail to yield a traceable and optimal representation space. An indispensable question arises: Can we concurrently address these limitations when reconstructing a feature space for a machine learning task? Our initial work took a pioneering step towards this challenge by introducing a novel self-optimizing framework. This framework leverages the power of three cascading reinforced agents to automatically select candidate features and operations for generating improved feature transformation combinations. Despite the impressive strides made, there was room for enhancing its effectiveness and generalization capability. In this extended journal version, we advance our initial work from two distinct yet interconnected perspectives: 1) We propose a refinement of the original framework, which integrates a graph-based state representation method to capture the feature interactions more effectively and develop different Q-learning strategies to alleviate Q-value overestimation further. 2) We utilize a new optimization technique (actor-critic) to train the entire self-optimizing framework in order to accelerate the model convergence and improve the feature transformation performance. Finally, to validate the improved effectiveness and generalization capability of our framework, we perform extensive experiments and conduct comprehensive analyses. These provide empirical evidence of the strides made in this journal version over the initial work, solidifying our framework’s standing as a substantial contribution to the field of automated feature transformation. To improve the reproducibility, we have released the associated code and data by the Github link https://github.com/coco11563/TKDD2023_code.

This article studies the problem of semi-supervised learning on graphs, which aims to incorporate ubiquitous unlabeled knowledge (e.g., graph topology, node attributes) with few-available labeled knowledge (e.g., node class) to alleviate the scarcity issue of supervised information on node classification. While promising results are achieved, existing works for this problem usually suffer from the poor balance of generalization and fitting ability due to the heavy reliance on labels or task-agnostic unsupervised information. To address the challenge, we propose a dual-channel framework for semi-supervised learning on G raphs via K nowledge T ransfer between independent supervised and unsupervised embedding spaces, namely, GKT. Specifically, we devise a dual-channel framework including a supervised model for learning the label probability of nodes and an unsupervised model for extracting information from massive unlabeled graph data. A knowledge transfer head is proposed to bridge the gap between the generalization and fitting capability of the two models. We use the unsupervised information to reconstruct batch-graphs to smooth the label probability distribution on the graphs to improve the generalization of prediction. We also adaptively adjust the reconstructed graphs by encouraging the label-related connections to solidify the fitting ability. Since the optimization of the supervised channel with knowledge transfer contains that of the unsupervised channel as a constraint and vice versa, we then propose a meta-learning-based method to solve the bi-level optimization problem, which avoids the negative transfer and further improves the model’s performance. Finally, extensive experiments validate the effectiveness of our proposed framework by comparing state-of-the-art algorithms.

Graph clustering has been a hot research topic and is widely used in many fields, such as community detection in social networks. Lots of works combining auto-encoder and graph neural networks have been applied to clustering tasks by utilizing node attributes and graph structure. These works usually assumed the inherent parameters (i.e., size and variance) of different clusters in the latent embedding space are homogeneous, and hence the assigned probability is monotonous over the Euclidean distance between node embeddings and centroids. Unfortunately, this assumption usually does not hold since the size and concentration of different clusters can be quite different, which limits the clustering accuracy. In addition, the node embeddings in deep graph clustering methods are usually L2 normalized so that it lies on the surface of a unit hyper-sphere. To solve this problem, we proposed D eep A daptive G raph C lustering via von Mises-Fisher distributions, namely DAGC. DAGC assumes the node embeddings H can be drawn from a von Mises-Fisher distribution and each cluster k is associated with cluster inherent parameters ρ k which includes cluster center μ and cluster cohesion degree κ. Then we adopt an EM-like approach (i.e., 𝒫( H | ρ ) and 𝒫( ρ | H ), respectively) to learn the embedding and cluster inherent parameters alternately. Specifically, with the node embeddings, we proposed to update the cluster centers in an attraction-repulsion manner to make the cluster centers more separable. And given the cluster inherent parameters, a likelihood-based loss is proposed to make node embeddings more concentrated around cluster centers. Thus, DAGC can simultaneously improve the intra-cluster compactness and inter-cluster heterogeneity. Finally, extensive experiments conducted on four benchmark datasets have demonstrated that the proposed DAGC consistently outperforms the state-of-the-art methods, especially on imbalanced datasets.

The goal of Feature Selection - comprising filter, wrapper, and embedded approaches - is to find the optimal feature subset for designated downstream tasks. Nevertheless, current feature selection methods are limited by: 1) the selection criteria of these methods are varied for different domains, leading them hard to be generalized; 2) the selection performance of these approaches drops significantly when processing high-dimensional feature space coupled with small sample size. In light of these challenges, we pose the question: can selected feature subsets be more robust, accurate, and input dimensionality agnostic? In this paper, we reformulate the feature selection problem as a deep differentiable optimization task and propose a new research perspective: conceptualizing discrete feature subsetting as continuous embedding space optimization. We introduce a novel and principled framework that encompasses a sequential encoder, an accuracy evaluator, a sequential decoder, and a gradient ascent optimizer. This comprehensive framework includes four important steps: preparation of features-accuracy training data, deep feature subset embedding, gradient-optimized search, and feature subset reconstruction. Specifically, we utilize reinforcement feature selection learning to generate diverse and high-quality training data and enhance generalization. By optimizing reconstruction and accuracy losses, we embed feature selection knowledge into a continuous space using an encodere-valuator-decoder model structure. We employ a gradient ascent search algorithm to find better embeddings in the learned embedding space. Furthermore, we reconstruct feature selection solutions using these embeddings and select the feature subset with the highest performance for downstream tasks as the optimal subset. Finally, extensive experimental results demonstrate the effectiveness of our proposed method, showcasing significant enhancements in feature selection robustness and accuracy. To improve the reproducibility of our research, we have released accompanying code and datasets by Dropbox.1.

In addressing the imbalanced issue of data within the realm of Natural Language Processing, text data augmentation methods have emerged as pivotal solutions. This data imbalance is prevalent in the research proposals submitted during the funding application process. Such imbalances, resulting from the varying popularity of disciplines or the emergence of interdisciplinary studies, significantly impede the precision of downstream topic models that deduce the affiliated disciplines of these proposals. At the data level, proposals penned by experts and scientists are inherently complex technological texts, replete with intricate terminologies, which augmenting such specialized text data poses unique challenges. At the system level, this, in turn, compromises the fairness of AI-assisted reviewer assignment systems, which raises a spotlight on solving this issue. This study leverages large language models (Llama V1) as data generators to augment research proposals categorized within intricate disciplinary hierarchies, aiming to rectify data imbalances and enhance the equity of expert assignments. We first sample within the hierarchical structure to find the under-represented class. Then we designed a prompt for keyword-based research proposal generation. Our experiments attests to the efficacy of the generated data, demonstrating that research proposals produced using the prompts can effectively address the aforementioned issues and generate high quality scientific text data, thus help the model overcome the imbalanced issue.

Automatic disease diagnosis from symptoms has attracted much attention in medical practices. It can assist doctors and medical practitioners in narrowing down disease candidates, reducing testing costs, improving diagnosis efficiency, and more importantly, saving human lives. Existing research has made significant progress in diagnosing disease but was limited by the gap between interpretability and accuracy. To fill this gap, in this paper, we propose a method called Reinforced Disease Diagnosis on Knowlege Graph (RDKG). Specifically, we first construct a knowledge graph containing all information from electronic medical records. To capture informative embeddings, we propose an enhanced knowledge graph embedding method that can embed information outside the knowledge graph into entity embedding. Then we transform the automatic disease diagnosis task into a Markov decision process on the knowledge graph. After that, we design a reinforcement learning method with a soft reward mechanism and a pruning strategy to solve the Markov decision process. We accomplish automated disease diagnosis by finding a path from symptoms to disease. The experimental results show that our model can effectively utilize heterogeneous information in the knowledge graph to complete the automatic disease diagnosis. Besides, our model demonstrates supreme performance in both accuracy and interpretability.

Feature transformation for AI is an essential task to boost the effectiveness and interpretability of machine learning (ML). Feature transformation aims to transform original data to identify an optimal feature space that enhances the performances of a downstream ML model. Existing studies either combines preprocessing, feature selection, and generation skills to empirically transform data, or automate feature transformation by machine intelligence, such as reinforcement learning. However, existing studies suffer from: 1) high-dimensional non-discriminative feature space; 2) inability to represent complex situational states; 3) inefficiency in integrating local and global feature information. To fill the research gap, we propose a novel group-wise cascading actor-critic perspective to develop the AI construct of automated feature transformation. Specifically, we formulate the feature transformation task as an iterative, nested process of feature generation and selection, where feature generation is to generate and add new features based on original features, and feature selection is to remove redundant features to control the size of feature space. Our proposed framework has three technical aims: 1) efficient generation; 2) effective policy learning; 3) accurate state perception. For an efficient generation, we develop a tailored feature clustering algorithm and accelerate generation by feature group-group crossing based generation. For effective policy learning, we propose a cascading actor-critic learning strategy to learn state-passing agents to select candidate feature groups and operations for fast feature generation. Such a strategy can effectively learn policies when the original feature size is large, along with exponentially growing feature generation action space, in which classic Q-value estimation methods fail. For accurate state perception of feature space, we develop a state comprehension method considering not only pointwise feature information but also pairwise feature-feature correlations. Finally, we present extensive experiments and case studies to illustrate 24.7% improvements in F1 scores compared with SOTAs and robustness in high-dimensional data.

With the explosive accumulation of urban geographic, mobile, and IoT service data, AI-assisted automated urban planning, with a goal of configuring land-uses, has become an emerging interdisciplinary topic for smart cities. Existing literature mostly views urban planning as a generative task from the perspective of generating land-use configuration images. Such perspective is limited by two issues: 1) hierarchical planning dependency across multi scales : there are hierarchical dependencies between region-level urban function configurations and block-level building configurations. 2) sequential planning dependency within a scale : when planning the buildings of a place, planning a shopping mall can impose constraints on planning subsequent Points of Interest (POIs). In response, we propose a new perspective of formulating urban planning as a hierarchical decision process. That is, given a target region with many geographic blocks, a machine planner firstly selects the optimized urban function portfolios, thereafter, sequentially selects the most appropriate POI for each block based on its urban functions and previously-placed POIs over planning steps. We reformulate this decision process into a hierarchical reinforcement learning task and develop a novel hierarchical reinforced urban planning framework. This framework includes two components: 1) In region-level configuration, we present an actor- critic based method to overcome the challenge of weak reward feedback in planning the urban functions of regions. 2) In block-level configuration, we propose a single-agent iterative POI allocation strategy to model dependencies between POIs and urban functions, and between current and previous POIs. Finally, we present extensive experimental results on real-world urban data to demonstrate the enhanced performances of the “planning as hierarchical decision process” perspective and the reinforced planning model.

With the development of natural language processing tech- niques(NLP), automatic diagnosis of eye diseases using ophthalmology electronic medical records (OEMR) has become possible. It aims to evaluate the condition of both eyes of a patient respectively, and we formulate it as a particular multi-label classification task in this paper. Although there are a few related studies in other diseases, automatic diagnosis of eye diseases exhibits unique characteristics. First, descriptions of both eyes are mixed up in OEMR documents, with both free text and templated asymptomatic descriptions, resulting in sparsity and clutter of information. Second, OEMR documents contain multiple parts of descriptions and have long document lengths. Third, it is critical to provide explainability to the disease diagnosis model. To overcome those challenges, we present an effective automatic eye disease diagnosis framework, NEEDED. In this framework, a preprocessing module is integrated to improve the density and quality of information. Then, we design a hierarchical transformer structure for learning the contextualized representations of each sentence in the OEMR document. For the diagnosis part, we propose an attention-based predictor that enables traceable diagnosis by obtaining disease-specific information. Experiments on the real dataset and comparison with several baseline models show the advantage and explainability of our framework.

The peer merit review of research proposals has been the major mechanism to decide grant awards. However, research proposals have become increasingly interdisciplinary. It has been a longstanding challenge to assign interdisciplinary proposals to appropriate reviewers so proposals are fairly evaluated. One of the critical steps in reviewer assignment is to generate accurate interdisciplinary topic labels for proposal-reviewer matching. Existing systems mainly collect topic labels manually generated by principle investigators. However, such human-reported labels can be non-accurate, incomplete, labor intensive, and time costly. What role can AI play in developing a fair and precise proposal reviewer assignment system? In this study, we collaborate with the National Science Foundation of China to address the task of automated interdisciplinary topic path detection. For this purpose, we develop a deep Hierarchical Interdisciplinary Research Proposal Classification Network (HIRPCN). Specifically, we first propose a hierarchical transformer to extract the textual semantic information of proposals. We then design an interdisciplinary graph and leverage GNNs to learn representations of each discipline in order to extract interdisciplinary knowledge. After extracting the semantic and interdisciplinary knowledge, we design a level-wise prediction component to fuse the two types of knowledge representations and detect interdisciplinary topic paths for each proposal. We conduct extensive experiments and expert evaluations on three real-world datasets to demonstrate the effectiveness of our proposed model.

Class-Incremental Learning (CIL) aims to train a reliable model with the streaming data, which emerges unknown classes sequentially. Different from traditional closed set learning, CIL has two main challenges: (1) Novel class detection. The initial training data only contains incomplete classes, and streaming test data will accept unknown classes. Therefore, the model needs to not only accurately classify known classes, but also effectively detect unknown classes; (2) Model expansion. After the novel classes are detected, the model needs to be updated without re-training using the entire previous data. However, traditional CIL methods have not fully considered these two challenges. To this end, we propose a Class-Incremental Learning without Forgetting (CILF) framework. In detail, CILF designs to regularize classification with decoupled prototype based loss, which can improve the intra-class and inter-class structure significantly, and acquire a compact embedding representation for novel class detection in result. Then, CILF employs a learnable curriculum clustering operator to estimate the number of semantic clusters via fine-tuning the learned network, in which curriculum operator can adaptively learn the embedding in self-taught form. Last, with the labeled streaming test data, CILF can update the network with robust regularization to mitigate the catastrophic forgetting.

In the era of big data, with the rapid increase in the number of texts, we have found ourselves submerged in the sea of texts. With the advent of the text summarization methodology, the time-consuming and energy-draining reading process of manual operations can be avoided by condensing and communicating the primary concept to the people. When generating sentences, the generator should comply with grammatical rules, such as lexical constraints. Leveraging lexical constraints in encoder-decoder models for the text summarization task has been extensively analyzed. Previous research add lexical constraints by enriching word embeddings in the encoding process. The model could be fed with more auxiliary knowledge. However, such lexical constraints cannot control the decoder, resulting in generating ungrammatical phrases. To address this issue, this paper proposes a grammar-aware text summarization method by incorporating Part-of-Speech (POS) constraints to guide the decoding process, along with auxiliary embeddings including POS, Lemmatization and Named Entity Recognition features in encoder. Detailed, the proposed POS-constrained AuxPOS model controls the generated word distribution during decoding according to the exact POS tag. Experimental results demonstrate that the proposed AuxPOS model can generate summaries that are more in concert with grammatical logic with high quality on the CNN/DailyMail, XSum and AESLC summarization datasets.

Human mobility forecasting can help us understand human movements, improve urban planning, and, ultimately, promote the development of livable, sustainable, and viable communities. While some efforts have been made for forecasting traffic or annotating trajectories, existing methods can be improved via simultaneously conducting human mobility forecasting and trajectory semantics augmentation. Along this line, in this paper, we provide a joint perspective of point processes and sequential embedding, in order to unify mobility arrival forecasting and trajectory semantics augmentation in a Hawkes-based long short-term memory (LSTM) method. Specifically, we first regard the traffic trajectories of a region as an arrival sequence according to the arrival time. Besides, we develop a method that exploits the mutual information of Hawkes processes and LSTM to model the arrival sequences of each region. Particularly, Hawkes processes predict the time and intensities of upcoming mobility arrivals; LSTM learns the embedding of arrivals, and annotates the arrival destinations and trip purposes; the mobility arrival intensities in Hawkes processes are influenced by the hidden states of LSTM. As applications, we exploit the proposed method to predict 3W (when, where, what) and discover functional regions. Finally, extensive experimental results with real-world traffic trajectory data demonstrate the enhanced performances of our method.

Apart from high-level computer vision tasks, deep learning has also made significant progress in low-level tasks, including single image dehazing. A well-detailed image looks realistic and natural with its clear edges and balanced colour. To achieve a clearer and vivid view, we exploit the role of edges and colours as a significant part of our proposed work. A progressive two-stage image dehazing network is presented to overcome the challenges of current image dehazing algorithms. The proposed image dehazing framework is divided into two steps; in the first stage, the multiscale image features of the encoder and decoder structure can be extracted. The second stage consists of the Color Correction Model (CCM), which retrieves balanced colour close to the ground truth. The encode-decoder network consists of a dense residual attention unit (DRAU) that comprises channel attention with pixel attention mechanisms. We have seen that weighted information and the haze difference is inconsistent across pixels without DRAU at the various channel-specific features. DRAU deals with different features and pixels unequally, which offers more versatility in handling knowledge of various types of detailed information. Our proposed two-stage network exceeds state-of-the-art algorithms in both visual and quantitative aspects. The findings are tested with the best-published peak signal-to-noise ratio metrics of 33.55–33.44 dB and SSIM 0.9619–0.9714 on SOTS indoor and outdoor test data sets.

Open set classification (OSC) tackles the problem of determining whether the data are in-class or out-of-class during inference, when only provided with a set of in-class examples at training time. Traditional OSC methods usually train discriminative or generative models with the owned in-class data, and then utilize the pre-trained models to classify test data directly. However, these methods always suffer from the embedding confusion problem, i.e., partial out-of-class instances are mixed with in-class ones of similar semantics, making it difficult to classify. To solve this problem, we unify semi-supervised learning to develop a novel OSC algorithm, S2OSC, which incorporates out-of-class instances filtering and model re-training in a transductive manner. In detail, given a pool of newly coming test data, S2OSC firstly filters the mostly distinct out-of-class instances using the pre-trained model, and annotates super-class for them. Then, S2OSC trains a holistic classification model by combing in-class and out-of-class labeled data with the remaining unlabeled test data in a semi-supervised paradigm. Furthermore, considering that data are usually in the streaming form in real applications, we extend S2OSC into an incremental update framework (I-S2OSC), and adopt a knowledge memory regularization to mitigate the catastrophic forgetting problem in incremental update. Despite the simplicity of proposed models, the experimental results show that S2OSC achieves state-of-the-art performance across a variety of OSC tasks, including 85.4% of F1 on CIFAR-10 with only 300 pseudo-labels. We also demonstrate how S2OSC can be expanded to incremental OSC setting effectively with streaming data.

To advance the development of science and technology, research proposals are submitted to open-court competitive programs developed by government agencies (e.g., NSF). Proposal classification is one of the most important tasks to achieve effective and fair review assignments. Proposal classification aims to classify a proposal into a length-variant sequence of labels. In this paper, we formulate the proposal classification problem into a hierarchical multi-label classification task. Although there are certain prior studies, proposal classification exhibit unique features: 1) the classification result of a proposal is in a hierarchical discipline structure with different levels of granularity; 2) proposals contain multiple types of documents; 3) domain experts can empirically provide partial labels that can be leveraged to improve task performances. In this paper, we focus on developing a new deep proposal classification framework to jointly model the three features. In particular, to sequentially generate labels, we leverage previously-generated labels to predict the label of next level; to integrate partial labels from experts, we use the embedding of these empirical partial labels to initialize the state of neural networks. Our model can automatically identify the best length of label sequence to stop next label prediction. Finally, we present extensive results to demonstrate that our method can jointly model partial labels, textual information, and semantic dependencies in label sequences, and, thus, achieve advanced performances.

Single Image Super-resolution (SISR) produces high-resolution images with fine spatial resolutions from a remotely sensed image with low spatial resolution. Recently, deep learning and generative adversarial networks (GANs) have made breakthroughs for the challenging task of single image super-resolution (SISR) . However, the generated image still suffers from undesirable artifacts such as the absence of texture-feature representation and high-frequency information. We propose a frequency domain-based spatio-temporal remote sensing single image super-resolution technique to reconstruct the HR image combined with generative adversarial networks (GANs) on various frequency bands (TWIST-GAN). We have introduced a new method incorporating Wavelet Transform (WT) characteristics and transferred generative adversarial network. The LR image has been split into various frequency bands by using the WT, whereas the transfer generative adversarial network predicts high-frequency components via a proposed architecture. Finally, the inverse transfer of wavelets produces a reconstructed image with super-resolution. The model is first trained on an external DIV2 K dataset and validated with the UC Merced Landsat remote sensing dataset and Set14 with each image size of 256 × 256. Following that, transferred GANs are used to process spatio-temporal remote sensing images in order to minimize computation cost differences and improve texture information. The findings are compared qualitatively and qualitatively with the current state-of-art approaches. In addition, we saved about 43% of the GPU memory during training and accelerated the execution of our simplified version by eliminating batch normalization layers.

While Graph Neural Network (GNN) has shown superiority in learning node representations of homogeneous graphs, leveraging GNN on heterogeneous graphs remains a challenging problem. The dominating reason is that GNN learns node representations by aggregating neighbors' information regardless of node types. Some work is proposed to alleviate such issue by exploiting relations or meta-path to sample neighbors with distinct categories, then use attention mechanism to learn different importance for different categories. However, one limitation is that the learned representations for different types of nodes should own different feature spaces, while all the above work still project node representations into one feature space. Moreover, after exploring massive heterogeneous graphs, we identify a fact that multiple nodes with the same type always connect to a node with another type, which reveals the many-to-one schema, a.k.a. the hierarchical tree structure. But all the above work cannot preserve such tree structure, since the exact multi-hop path correlation from neighbors to the target node would be erased through aggregation. Therefore, to overcome the limitations of the literature, we propose T-GNN, a tree structure-aware graph neural network model for graph representation learning. Specifically, the proposed T-GNN consists of two modules: (1) the integrated hierarchical aggregation module and (2) the relational metric learning module. The integrated hierarchical aggregation module aims to preserve the tree structure by combining GNN with gated recurrent unit to integrate the hierarchical and sequential neighborhood information on the tree structure to node representations. The relational metric learning module aims to preserve the heterogeneity by embedding each type of nodes into a type-specific space with distinct distribution based on similarity metrics. In this way, our proposed T-GNN is capable of simultaneously preserving the heterogeneity and the tree structure inherent in heterogeneous graphs. Finally, we conduct extensive experiments to show the outstanding performance of T-GNN in tasks of node clustering and classification, inductive node clustering and classification, and link prediction.

People share same names in real world. When a digital library user searches for an author name, he may see a mixture of publications by different authors who have the same name. Making distinctions between them is an important prerequisite to improve the quality of services and contents in digital libraries. The general task of author disambiguation is to associate publications which belong to an identical name or names with highly similar spellings to different people entities. In recent years, many researches have been conducted to solve this challenging task. However, some works rely heavily on external knowledge bases and manually annotated data. Some unsupervised learning based works require complex feature engineering. In this paper, we propose a novel and efficient author disambiguation framework which needs no labeled data. We first construct a publication heterogeneous network for each ambiguous name. Then, we use our proposed heterogeneous graph convolutional network embedding method that encodes both graph structure and node attribute information to learn publication representations. After that, we propose a graph enhanced clustering method for name disambiguation that can greatly accelerate the clustering process and need not require the number of distinct persons. Our framework can be continually retrained and applied on incremental disambiguation task when new publications are put in. Experimental results on two datasets show that our framework clearly performs better than several state-of-the-art methods for author disambiguation.

What is the purpose of a trip? What are the unique human mobility patterns and spatial contexts in or near the pickup points and delivery points of trajectories for a specific trip purpose? Many prior studies have modeled human mobility patterns in urban regions; however, these analytics mainly focus on interpreting the semantic meanings of geographic topics at an aggregate level. Given the lack of information about human activities at pick-up and dropoff points, it is challenging to convert the prior studies into effective tools for inferring trip purposes. To address this challenge, in this article, we study large-scale taxi trajectories from an unsupervised perspective in light of the following observations. First, the POI configurations of origin and destination regions closely relate to the urban functionality of these regions and further indicate various human activities. Second, with respect to the functionality of neighborhood environments, trip purposes can be discerned from the transitions between regions with different functionality at particular time periods. Along these lines, we develop a general probabilistic framework for spotting trip purposes from massive taxi GPS trajectories. Specifically, we first augment the origin and destination regions of trajectories by attaching neighborhood POIs. Then, we introduce a latent factor, POI Topic , to represent the mixed functionality of the regions, such that each origin or destination point in the city can be modeled as a mixture over POI Topics. In addition, considering the transitions from origins to destinations at specific time periods, the trip time is generated collaboratively from the pairwise POI Topics at both ends of the O-D pairs, constituting POI Links , and hence the trip purpose can be explained semantically by the POI Links. Finally, we present extensive experiments with the real-world data of New York City to demonstrate the effectiveness of our proposed method for spotting trip purposes, and moreover, the model is validated to perform well in predicting the destinations and trip time among all the baseline methods.

The problem of software artifact retrieval has the goal to effectively locate software artifacts, such as a piece of source code, in a large code repository. This problem has been traditionally addressed through the textual query. In other words, information retrieval techniques will be exploited based on the textual similarity between queries and textual representation of software artifacts, which is generated by collecting words from comments, identifiers, and descriptions of programs. However, in addition to these semantic information, there are rich information embedded in source codes themselves. These source codes, if analyzed properly, can be a rich source for enhancing the efforts of software artifact retrieval. To this end, in this paper, we develop a feature extraction method on source codes. Specifically, this method can capture both the inherent information in the source codes and the semantic information hidden in the comments, descriptions, and identifiers of the source codes. Moreover, we design a heterogeneous metric learning approach, which allows to integrate code features and text features into the same latent semantic space. This, in turn, can help to measure the artifact similarity by exploiting the joint power of both code and text features. Finally, extensive experiments on real-world data show that the proposed method can help to improve the performances of software artifact retrieval with a significant margin.

Trajectory data is becoming more and more popular nowadays and extensive studies have been conducted on trajectory data. One important research direction about trajectory data is the anomaly detection which is to find all anomalies based on trajectory patterns in a road network. In this paper, we introduce a road segment-based anomaly detection problem, which is to detect the abnormal road segments each of which has its “real” traffic deviating from its “expected” traffic and to infer the major causes of anomalies on the road network. First, a deviation-based method is proposed to quantify the anomaly of reach road segment. Second, based on the observation that one anomaly from a road segment can trigger other anomalies from the road segments nearby, a diffusion-based method based on a heat diffusion model is proposed to infer the major causes of anomalies on the whole road network. To validate our methods, we conduct intensive experiments on a large real-world GPS dataset of about 23,000 taxis in Shenzhen, China to demonstrate the performance of our algorithms.