The human visual system is capable of not only recognizing individual objects but also comprehending the contextual relationship between them in real-world scenarios, making it highly advantageous for object detection. However, in practical applications, such contextual information is often not available. Previous attempts to compensate for this by utilizing cross-modal data such as language and statistics to obtain contextual priors have been deemed sub-optimal due to a semantic gap. To overcome this challenge, we present a seamless integration of context into an object detector through Knowledge Distillation. Our approach intuitively represents context as a knowledge graph, describing the relative location and semantic relevance of different visual concepts. Leveraging recent advancements in graph representation learning with Transformer, we exploit the contextual information among objects using edge encoding and graph attention. Specifically, each image region propagates and aggregates the representation from its highly similar neighbors to form the knowledge graph in the Transformer encoder. Extensive experiments and a thorough ablation study conducted on challenging benchmarks MS-COCO, Pascal VOC and LVIS demonstrate the superiority of our method.

A patient's previous history of malaria plays an important role in malaria prediction based on symptoms. Malaria patients who have previously suffered from malaria are highly likely to develop different symptoms from those without a previous history of malaria. To predict the malaria presence for both patients with and without a previous history of malaria, we build two separate malaria classifiers based on two different sets of symptoms using four machine learning techniques including neural networks (NNs), logistic regression (LR), support vector machines (SVMs) and k-nearest neighbors. These malaria classifiers are built using medical records collected from patients suffering from malaria and other febrile diseases. Extensive experiments conducted show that the two NN classifiers slightly outperform the other classifiers. The NN classifier for patients with a previous history of malaria achieves excellent performance for accuracy, recall and F1-score with 95.76%, 95.41% and 95.76% respectively. The LR classifier outperforms the other three classifiers for precision with 97.14%. The NN classifier for patients without a previous history of malaria also achieves superior performance in accuracy and precision with 88.48% and 87.34% respectively. The SVM classifier outperforms other classifiers in terms of recall and F1-score with 94.62% and 88.04% respectively. With high recall rates, the four classifiers are suitable for symptom-based malaria prediction. These results justify the development of two separate malaria classifiers and provide valuable insights into building symptom-based malaria classifiers based on patients' previous history of malaria.

Personalized outfit recommendation, which aims to recommend the outfits to a given user according to his/her preference, has gained increasing research attention due to its economic value. Nevertheless, the majority of existing methods mainly focus on improving the recommendation effectiveness, while overlooking the recommendation efficiency. Inspired by this, we devise a novel bi-directional heterogeneous graph hashing scheme, called BiHGH, towards efficient personalized outfit recommendation. In particular, this scheme consists of three key components: heterogeneous graph node initialization, bi-directional sequential graph convolution, and hash code learning. We first unify four types of entities (i.e., users, outfits, items, and attributes) and their relations via a heterogeneous four-partite graph. To perform graph learning, we then creatively devise a bi-directional graph convolution algorithm to sequentially transfer knowledge via repeating upwards and downwards convolution, whereby we divide the four-partite graph into three subgraphs and each subgraph only involves two adjacent entity types. We ultimately adopt the bayesian personalized ranking loss for the user preference learning and design the dual similarity preserving regularization to prevent the information loss during hash learning. Extensive experiments on the benchmark dataset demonstrate the superiority of BiHGH.

Fashion Compatibility Modeling (FCM) is a new yet challenging task, which aims to automatically access the matching degree among a set of complementary items. Most of existing methods evaluate the fashion compatibility from the common perspective, but overlook the user's personal preference. Inspired by this, a few pioneers study the Personalized Fashion Compatibility Modeling (PFCM). Despite their significance, these PFCM methods mainly concentrate on the user and item entities, as well as their interactions, but ignore the attribute entities, which contain rich semantics. To address this problem, we propose to fully explore the related entities and their relations involved in PFCM to boost the PFCM performance. This is, however, non-trivial due to the heterogeneous contents of different entities, embeddings for new users, and various high-order relations. Towards these ends, we present a novel metapath-guided personalized fashion compatibility modeling, dubbed as MG-PFCM. In particular, we creatively build a heterogeneous graph to unify the three types of entities (i.e., users, items, and attributes) and their relations (i.e., user-item interactions, item-item matching relations, and item-attribute association relations). Thereafter, we design a multi-modal content-oriented user embedding module to learn user representations by inheriting the contents of their interacted items. Meanwhile, we define the user-oriented and item-oriented metapaths, and perform the metapath-guided heterogeneous graph learning to enhance the user and item embeddings. In addition, we introduce the contrastive regularization to improve the model performance. We conduct extensive experiments on the real-world benchmark dataset, which verifies the superiority of our proposed scheme over several cutting-edge baselines. As a byproduct, we have released our source codes to benefit other researchers.

Fashion Compatibility Modeling (FCM), which aims to automatically evaluate whether a given set of fashion items makes a compatible outfit, has attracted increasing research attention. Recent studies have demonstrated the benefits of conducting the item representation disentanglement towards FCM. Although these efforts have achieved prominent progress, they still perform unsatisfactorily, as they mainly investigate the visual content of fashion items, while overlooking the semantic attributes of items (e.g., color and pattern), which could largely boost the model performance and interpretability. To address this issue, we propose to comprehensively explore the visual content and attributes of fashion items towards FCM. This problem is non-trivial considering the following challenges: a) how to utilize the irregular attribute labels of items to partially supervise the attribute-level representation learning of fashion items; b) how to ensure the intact disentanglement of attribute-level representations; and c) how to effectively sew the multiple granulairites (i.e, coarse-grained item-level and fine-grained attribute-level) information to enable performance improvement and interpretability. To address these challenges, in this work, we present a partially supervised outfit compatibility modeling scheme (PS-OCM). In particular, we first devise a partially supervised attribute-level embedding learning component to disentangle the fine-grained attribute embeddings from the entire visual feature of each item. We then introduce a disentangled completeness regularizer to prevent the information loss during disentanglement. Thereafter, we design a hierarchical graph convolutional network, which seamlessly integrates the attribute- and item-level compatibility modeling, and enables the explainable compatibility reasoning. Extensive experiments on the real-world dataset demonstrate that our PS-OCM significantly outperforms the state-of-the-art baselines. We have released our source codes and well-trained models to benefit other researchers (https://site2750.wixsite.com/ps-ocm).

Existing methods towards outfit compatibility modeling seldom explicitly consider multimodal correlations. In this work, we explore the consistent and complementary correlations for better compatibility modeling. This is, however, non-trivial due to the following challenges: 1) how to separate and model these two kinds of correlations; 2) how to leverage the derived complementary cues to strengthen the text and vision-oriented representations of the given item; and 3) how to reinforce the compatibility modeling with text and vision-oriented representations. To address these challenges, we present a comprehensive multimodal outfit compatibility modeling scheme. It first nonlinearly projects each modality into separable consistent and complementary spaces via multi-layer perceptron, and then models the consistent and complementary correlations between two modalities by parallel and orthogonal regularization. Thereafter, we strengthen the visual and textual representation of items with complementary information, and further induct both the text-oriented and vision- oriented outfit compatibility modeling. We ultimately employ the mutual learning strategy to reinforce the final performance of compatibility modeling. Extensive experiments demonstrate the superiority of our scheme.

Selecting non-ferrous scrap metal suppliers involves evaluating each supplier and its available scrap metals in terms of quantitative and qualitative criteria in each selection round. This paper presents a new approach for developing and selecting fuzzy multi-criteria decision making (MCDM) methods to address the non-ferrous scrap metal supplier selection problem with fuzzy data. The approach develops different fuzzy MCDM methods by combining three normalization processes and three aggregation processes commonly used in MCDM research. In particular, a specific defuzzification process is used to reflect the decision maker's risk attitude about the current global commodity market. A validation procedure using fuzzy c-means clustering is applied to select among inconsistent ranking results produced by different fuzzy MCDM methods. An empirical study on a non-ferrous scrap metal company in China is conducted to illustrate how fuzzy MCDM methods are developed and selected.

Most existing object detection models are restricted to detecting objects from previously seen categories, an approach that tends to become infeasible for rare or novel concepts. Accordingly, in this paper, we explore object detection in the context of zero-shot learning, i.e., Zero-Shot Object Detection (ZSD), to concurrently recognize and localize objects from novel concepts. Existing ZSD algorithms are typically based on a simple mapping-transfer strategy that is susceptible to the domain shift problem. To resolve this problem, we propose a novel Semantics-Preserving Graph Propagation model for ZSD based on Graph Convolutional Networks (GCN). More specifically, we employ a graph construction module to flexibly build category graphs by incorporating diverse correlations between category nodes; this is followed by two semantics preserving modules that enhance both category and region representations through a multi-step graph propagation process. Compared to existing mapping-transfer based methods, both the semantic description and semantic structural knowledge exhibited in prior category graphs can be effectively leveraged to boost the generalization capability of the learned projection function via knowledge transfer, thereby providing a solution to the domain shift problem. Experiments on existing seen/unseen splits of three popular object detection datasets demonstrate that the proposed approach performs favorably against state-of-the-art ZSD methods.

Neural Networks (NNs) are non-linear models and are widely used to model complex relationships, thus being well suited to formulate the product design process for matching design form elements to consumers’ affective preferences. In this paper, we construct 36 deep NN models, using one to four hidden layers with three different dropout ratios and three widely used rules for determining the number of neurons in the hidden layer(s). As a result of extensive experiments, the NN model using one hidden layer with 140 hidden neurons has the highest predicting accuracy rate (80%) and is used to help product designers determine the optimal form combination for new fragrance bottle design.

Existing approaches to Chinese semantic role labeling (SRL) mainly adopt deep long short-term memory (LSTM) neural networks to address the long-term dependencies problem. However, deep LSTM networks cannot address the vanishing gradient problem properly. In addition, the complexity of the Chinese language, as a hieroglyphic language, decreases the performance of traditional SRL approaches to Chinese SRL. To address these problems, this paper proposes a new approach with a deep bidirectional highway LSTM network. The performance of the proposed approach is further improved by introducing the conditional random fields (CRFs) constraints and part-of-speech (POS) feature since POS tags are the classes of formal equivalents of words in linguistics. The experimental results on the commonly used Chinese Proposition Bank dataset show that the proposed approach outperforms existing approaches. With an easily acquired and reliable POS feature for practical applications, the proposed approach substantially improves Chinese SRL.

This paper proposes a new iterative fuzzy clustering (IFC) algorithm to impute missing values of datasets. The information provided by fuzzy clustering is used to update the imputed values through iterations. The performance of the IFC algorithm is examined by conducting experiments on three commonly used datasets and a case study on a city mobility database. Experimental results show that the IFC algorithm not only works well for datasets with a small number of missing values but also provides an effective imputation result for datasets where the proportion of missing data is high.

This paper develops a new method called bitechnique for order preference by similarity to ideal solution (Bi-TOPSIS) to address the multicriteria decision making problems involving interrelated criteria using bipolar measurement with negative values. The Bi-TOPSIS method incorporates the capability of the bi-capacity technique into the TOPSIS to address two important issues: 1) how to measure the interactions between criteria and 2) how to aggregate values measured on a bipolar scale. In practical applications, this method allows the use of benchmarks to demarcate “good” from “bad” performance, thus enhancing the interpretability of the evaluation results. To examine its effectiveness, an empirical study on the evaluation of city social sustainability is conducted. The outcome of the study demonstrates the feasibility and applicability of the new Bi-TOPSIS method.

Artificial neural networks (ANNs) have been applied successfully in a wide range of fields due to its effective learning ability. In this paper, we propose a grey relational analysis (GRA) based ANN model that can be used to build a design decision support database for facilitating the product design process and matching specific consumers' preferences. The result of an empirical application and a comparative study on fragrance bottle form design shows that the ANN models outperform the grey prediction models, indicating that the ANN technique is promising to help product designers design a new product that best meets consumers' needs.

In this paper, we develop a new Fuzzy Cognitive Map (FCM) learning method using the imperialistic competitive learning algorithm (ICLA). An FCM seems like a fuzzy signed directed graph with feedback, and models complex systems as a collection of concepts and causal relations between concepts. Conventional FCMs are mainly constructed by human experts who have experience in the specific problem domain. However, large problems need automated methods. We develop an automated method for FCM construction inspired by the socio-political behavior of countries as imperialists with colonies. In the real world imperialists extend their territories and change the socio attributes of their colonies. The ICLA is an evolutionary algorithm and simulates this behavior. We explain the algorithm for FCM learning and demonstrate its performance advantages through synthetic and real data of water demand. The results of the new algorithm were compared to that of a genetic algorithm, which is the most commonly used and well-known FCM learning algorithm.

This paper presents a fuzzy multiattribute decision making (MADM) approach to evaluating airport performance in terms of the airport operator, the passenger, and the airline dimensions. Assessments for quantitative measures are represented by crisp values, and assessments for qualitative measures and attribute weights are represented by triangular fuzzy numbers, obtained via surveys using linguistic terms. An effective algorithm is developed by using the α-cut concept for incorporating the decision-maker's confidence level about respondents' fuzzy assessments on performance ratings and attribute weights. Based on the concept of the degree of optimality, the algorithm obtains a quantitative performance index for each airport evaluated, relative to other comparable airports with respect to each of three evaluation dimensions individually and all three evaluation dimensions as a whole. The performance index obtained would help individual airports understand the comparative level of their performance on measures relating to operational management, airport facilities, and passenger service quality. An empirical study of performance evaluation of 11 Asia-Pacific major international airports is conducted to illustrate how the approach works.

This paper presents a simulation-based study to evaluate the performance of 12 defuzzification-based approaches for solving the general fuzzy multiattribute decision-making (MADM) problem requiring cardinal ranking of decision alternatives. These approaches are generated based on six defuzzification methods in conjunction with the simple additive weighting (SAW) method and the technique for order preference by similarity to the ideal solution method. The consistency and effectiveness of these approaches are examined in terms of four new objective performance measures, which are based on five evaluation indexes. The simulation result shows that the approaches, which are capable of using all the available information on fuzzy numbers effectively in the defuzzification process, produce more consistent ranking outcomes. In particular, the SAW method with the degree of dominance defuzzification is proved to be the overall best performed approach, which is followed by the SAW method with the area center defuzzification. These findings are of practical significance in real-world settings where the selection of the defuzzification-based approaches is required in solving the general fuzzy MADM problems under specific decision contexts.

An emerging requirement to sift through the increasing flood of multilingual text available electronically has led to the pressing demand for effective multilingual information filtering. In this paper, a content-based approach to multilingual information filtering is proposed. This approach is capable of screening and evaluating multilingual documents based on their semantic content. As such, relevant multilingual documents are disseminated according to their corresponding themes/topics to facilitate both efficient and effective content-based information access. The objective of alleviating users' burden of information overload is thus achieved. This approach is realized by incorporating fuzzy clustering and fuzzy inference techniques.

Fuzzy multiattribute decision making (MADM) has been widely used in ranking a finite number of decision alternatives characterized by fuzzy assessments with respect to multiple criteria. The overall utility of an alternative is obtained by aggregating the criteria weights and alternatives ratings, on which the ranking is based. However, different aggregation techniques often produce inconsistent outcomes for the same problem. This paper presents a validation procedure for examining the validity of fuzzy MADM methods, in order to assist in selecting a valid outcome for a given selection problem. The procedure uses fuzzy clustering as a basis to compare with the results from fuzzy MADM methods. A case study of the scrap metal buyer selection is conducted to demonstrate the effectiveness of the procedure.

Resource-constrained Project Scheduling (RCPS) is one of the NP-hard classes because of the complexity of their combinatorial nature. Any exact algorithms are difficult to handle such a problem. In addition, uncertainty also often exists in many projects in the process of the estimate of activity durations. This paper presents the framework of a fuzzy genetic algorithm combined with tabu mechanism for fuzzy RCPS in order to obtain an approximate optimal solution. This approach can handle both fuzzy and crisp numbers that commonly coexist in many realistic RCPS problems. To demonstrate the algorithm developed, a real case of an overhaul schedule is considered as a fuzzy RCPS problem.

This paper formulates the performance evaluation of service quality among a set of competing service companies as a fuzzy multiattribute decision making problem, and presents a preference-based approach for obtaining an overall service performance index for each company. The data required are obtained by customer surveys. Subjective assessments of service attribute weights and performance ratings of individual companies are represented by fuzzy numbers, using linguistic terms. Respondents have the option to define the value range of linguistic terms for use in their assessments. Based on the concept of the degree of optimality, the preference-based approach uses the /spl alpha/-cut concept to allow the incorporation of the decision-maker's confidence level on respondents' fuzzy assessments. For a given confidence level, the decision-maker's attitude toward risk can be incorporated to reflect the decision maker's preference on crisp value intervals derived from fuzzy assessments.

In project management, uncertainty often exists in relation to the estimation of activity durations. Fuzzy set theory is an effective way of solving the uncertainty caused by a lack of sufficient precise information. To schedule a project under conditions of limited resources, a genetic algorithm combined with a tabu mechanism is developed in order to search effectively and efficiently for a global-optimal or near-optimal solution. The fuzzy algorithm developed has been integrated into a computer-aided system. An experiment has demonstrated that the system is beneficial for solving such NP-hard problems when the project size increases.

Ranking fuzzy numbers plays a critical role in decision analysis under a fuzzy environment. Existing fuzzy ranking methods may not always be suitable for practical decision problems of large size, due to counter-intuitive ranking outcomes produced or considerable computational effort required. This paper presents an effective approach to address the fuzzy ranking problem of practical size. The proposed approach combines the merits of two prominent concepts individually used in the literature: the fuzzy reference set and the degree of dominance. As such, the decisive information embedded in the set of fuzzy numbers to be ranked is sensibly used, and satisfactory ranking outcomes are always achieved. The approach is computationally simple and its underlying concepts are logically sound and comprehensible. A comparative study is conducted on all benchmark cases used in the literature to examine its performance on rationality and discriminatory ability. The comparison result shows that the approach compares favorably with comparable methods examined.