Computational Intelligence最新文献

Session-based recommendation that uses sequence of items clicked by anonymous users to make recommendations has drawn the attention of many researchers, and a lot of approaches have been proposed. However, there are still problems that have not been well addressed: (1) Time information is either ignored or exploited with a fixed time span and granularity, which fails to understand the personalized interest transfer pattern of users with different clicking speeds; (2) Category information is either omitted or considered independent of the items, which defies the fact that the relationships between categories and items are helpful for the recommendation. To solve these problems, we propose a new session-based recommendation method, TCSR (self-attention with time and category for session-based recommendation). TCSR uses a non-linear normalized time embedding to perceive user interest transfer patterns on variable granularity and employs a heterogeneous SAN to make full use of both items and categories. Moreover, a cross-recommendation unit is adapted to adjust recommendations on the item and category sides. Extensive experiments on four real datasets show that TCSR significantly outperforms state-of-the-art approaches.

In intelligent transportation systems (ITS), achieving accurate travel time estimation (TTE) is paramount, much like route planning. Precisely predicting travel time across different urban areas is vital, and an essential requirement for these privileges is having fine-grained knowledge of the city. In contrast to prior studies that are restricted to coarse-grained data, we broaden the scope of traffic flow forecasting to fine granularity, which provokes explicit challenges: (1) the prevalence of inter-grid transitions within fine-grained data introduces complexity in capturing spatial dependencies among grid cells on a global scale. (2) stemming from dynamic temporal dependencies. To address these challenges, we propose the multi-scaling hybrid model (MSHM) as a novel approach. Initially, a multi-directional convolutional layer is first used to acquire high-level depictions for each cell to retrieve the semantic attributes of the road network from local and global aspects. Next, we incorporate the characteristics of the road network and coarse-grained flow features to regularize the local and global spatial distribution modeling of road-relative traffic flow using an enhanced deep super-resolution (EDSR) technique. Benefiting from the EDSR method, our approach can generate high-quality fine-grained traffic flow maps. Furthermore, to continuously provide accurate TTE over time by leveraging well-designed multi-scale feature modeling, we incorporate a multi-scale feature expression of each road segment, capturing intricate details and important features at different scales to optimize the TTE. We conducted comprehensive trials on two real-world datasets, BJTaxi and NYCTaxi, aiming to achieve superior results compared to baseline methods.

The integration of different Mobile Edge Computing (MEC) applications has significantly enhanced the realm of security and surveillance, with Human Activity Recognition (HAR) standing out as a crucial application. The diverse sensors found in smartphones have made it convenient for monitoring applications to gather and analyze data, rendering them valuable for HAR purposes. Moreover, MEC can be employed to automate surveillance, allowing intelligent monitoring of restricted areas to identify and respond to unwanted or suspicious activities. This research develops a system using motion sensors in smartphones to identify unusual human activities. People's smartphones were employed to monitor both suspicious and regular activities. Information was collected for various actions categorized as either suspicious or regular. When a person performs a certain action, the smartphone records a series of sensory data, analyse important patterns from the basic data, and then determines what the person is doing by combining information from different sensors. To prepare the data, information from different sensors was aligned to a shared timeline. In this study, we used a sliding window approach on synchronized data to feed sequences into LSTM and CNN models. These models, which include initial layers of LSTM and CNN, automatically find important patterns in the order of human activities. We combined SVM with the features extracted by the shallow Neural Network to make a mixed model that predicts suspicious activities. Lastly, we compared LSTM, CNN, and our new shallow mixed neural network using a new real-time dataset. The mixed model of CNN and SVM achieved an accuracy of 94.43%. Additionally, the sliding window method's effectiveness was confirmed with a 4.28% improvement in accuracy.

Protein sequence classification needs to be performed quickly and accurately to progress bioinformatics advancements and the production of pharmaceutical products. Extensive comparisons between large databases of known proteins and unknown sequences are necessary in traditional protein classification methods, which can be time-consuming. This labour-intensive and slow manual matching and classification method depends on functional and biological commonalities. Protein classification is one of the many fields in which deep learning has recently revolutionized. The data on proteins are organized hierarchically and sequentially, and the most advanced algorithms, such as Deep Family-based Method (DeepFam) and Protein Convolutional Neural Network (ProtCNN), have shown promising results in classifying proteins into relative groups. On the other hand, these methods frequently refuse to acknowledge this fact. We propose a novel hybrid model called ProteinCNN-BLSTM to overcome these particular challenges. To produce more accurate protein sequence classification, it combines the techniques of amino acid embedding with bidirectional long short-term memory (BLSTM) and convolutional neural networks (CNNs). The CNN component is the most effective at capturing local features, while the BLSTM component is the most capable of modeling long-term dependencies across protein sequences. Through the process of amino acid embedding, sequences of proteins are transformed into numeric vectors, which significantly improves the precision of prediction and the representation of features. Using the standard protein samples PDB-14189 and PDB-2272, we analyzed the proposed ProteinCNN-BLSTM model and the existing deep-learning models. Compared to the existing models, such as CNN, LSTM, GCNs, CNN-LSTM, RNNs, GCN-RNN, DeepFam, and ProtCNN, the proposed model performed more accurately and better than the existing models.

Deep learning models have overcome traditional machine learning techniques for text classification domains in the field of natural language processing (NLP). Since, NLP is a branch of machine learning, used for interpreting language, classifying text of interest, and the same can be applied to analyse the medical clinical electronic health records. Medical text consists of lot of rich data which can altogether provide a good insight, by determining patterns from the clinical text data. In this paper, bidirectional-long short-term memory (Bi-LSTM), bi-LSTM attention and bidirectional encoder representations from transformers (BERT) base models are used to classify the text which are of privacy concern to a person and which should be extracted and can be tagged as sensitive. This text data which we might think not of privacy concern would majorly reveal a lot about the patient's integrity and personal life. Clinical data not only have patient demographic data but lot of hidden data which might go unseen and thus could arise privacy issues. Bi-LSTM with attention layer is also added on top to realize the importance of critical words which will be of great importance in terms of classification, we are able to achieve accuracy of about 92%. About 206,926 sentences are used out of which 80% are used for training and rest for testing we get accuracy of 90% approx. with Bi-LSTM alone. The same set of datasets is used for BERT model with accuracy of 93% approx.

In the field of single image super-resolution, the prevalent use of convolutional neural networks (CNN) typically assumes a simplistic bicubic downsampling model for image degradation. This assumption misaligns with the complex degradation processes encountered in medical imaging, leading to a performance gap when these algorithms are applied to real medical scenarios. Addressing this critical discrepancy, our study introduces a novel degradation comparative learning framework meticulously designed for the nuanced degradation characteristics of medical images within the Internet of Medical Things (IoMT). Unlike traditional CNN-based super-resolution approaches that homogeneously process image channels, our method acknowledges and leverages the disparity in informational content across channels. We present a blind image super-resolution technique, underpinned by edge reconstruction and an innovative image feature supplement module. This approach not only preserves but enriches texture details, crucial for the accurate analysis of medical images in the IoMT. Comparative evaluations of our model against existing blind super-resolution methods, utilizing both natural image testing datasets and medical images, demonstrate its superior performance. Notably, our approach exhibits remarkable proficiency in stably restoring various degraded super-resolution images, a critical requirement in the IoMT context. Experimental results demonstrate that our method is superior to the current state-of-the-art methods, marking a significant advancement in the field of medical image super-resolution.

Brain tumor segmentation from MRI is a challenging process that has positive ups and downs. The most crucial step for detection and treatment to save the patient's life is earlier diagnosis and classification of brain tumor (BT) with higher accuracy prediction. One of the deadliest cancers, malignant brain tumors is now the main cause of cancer-related death due to their extreme severity. To evaluate the tumors and help patients receive the appropriate treatment according to their classifications, it is essential to have a thorough understanding of brain diseases, such as classifying BT. In order to resolve the problem of low segmentation accuracy caused by an imbalance of model design and sample category in the process of brain tumor segmentation. In this research work, Multi-Dimensional Cascades Neural Network (MDCNet) is developed for multi-class BT classification. It is divided into two steps. In stage 1, an enhanced shallow-layer 3D locality net is used to conduct BT localization and rough segmentation on the preprocessed MRIs. It is also advised to use a unique circular inference module and parameter Dice loss to lower the uncertain probability and false positive border locations. In step 2, in order to compensate for mistakes and lost spatial information of a single view, morphological traits are investigated using a multi-view 2.5D net composed of three 2D refinement subnetworks. The suggested method outperforms the traditional model in segmentation, yielding an accuracy of 99.67%, 98.16%, and 99.76% for the three distinct datasets.

Modelling one-to-many type mappings in problems with a temporal component can be challenging. Backpropagation is not applicable to networks that perform discrete sampling and is also susceptible to gradient instabilities, especially when applied to longer sequences. In this paper, we propose two recurrent neural network architectures that leverage stochastic units and mixture models, and are trained with target propagation. We demonstrate that these networks can model complex conditional probability distributions, outperform backpropagation-trained alternatives, and do not rapidly degrade with increased time horizons. Our main contributions consist of the design and evaluation of the architectures that enable the networks to solve multi-model problems with a temporal dimension. This also includes the extension of the target propagation through time algorithm to handle stochastic neurons. The use of target propagation provides an additional computational advantage, which enables the network to handle time horizons that are substantially longer compared to networks fitted using backpropagation.

Question-driven automatic text summarization is a popular technique to produce concise and informative answers to specific questions using a document collection. Both query-based and question-driven summarization may not produce reliable summaries nor contain relevant information if they do not take advantage of extractive and abstractive summarization mechanisms to improve performance. In this article, we propose a novel extractive and abstractive hybrid framework designed for question-driven automatic text summarization. The framework consists of complimentary modules that work together to generate an effective summary: (1) discovering appropriate non-redundant sentences as plausible answers using an open-domain multi-hop question answering system based on a convolutional neural network, multi-head attention mechanism and reasoning process; and (2) a novel paraphrasing generative adversarial network model based on transformers rewrites the extracted sentences in an abstractive setup. Experiments show this framework results in more reliable abstractive summary than competing methods. We have performed extensive experiments on public datasets, and the results show our model can outperform many question-driven and query-based baseline methods (an R1, R2, RL increase of 6%–7% for over the next highest baseline).

There is a growing trend of using artificial intelligence, particularly deep learning algorithms, in medical diagnostics, revolutionizing healthcare by improving efficiency, accuracy, and patient outcomes. However, the use of artificial intelligence in medical diagnostics comes with the critical need to explain the reasoning behind artificial intelligence-based predictions and ensure transparency in decision-making. Explainable artificial intelligence has emerged as a crucial research area to address the need for transparency and interpretability in medical diagnostics. Explainable artificial intelligence techniques aim to provide insights into the decision-making process of artificial intelligence systems, enabling clinicians to understand the factors the algorithms consider in reaching their predictions. This paper presents a detailed review of saliency-based (visual) methods, such as class activation methods, which have gained popularity in medical imaging as they provide visual explanations by highlighting the regions of an image most influential in the artificial intelligence's decision. We also present the literature on non-visual methods, but the focus will be on visual methods. We also use the existing literature to experiment with infrared breast images for detecting breast cancer. Towards the end of this paper, we also propose an “attention guided Grad-CAM” that enhances the visualizations for explainable artificial intelligence. The existing literature shows that explainable artificial intelligence techniques are not explored in the context of infrared medical images and opens up a wide range of opportunities for further research to make clinical thermography into assistive technology for the medical community.