Cognitive Computation最新文献

Abstract

This article discusses the fixed-time synchronization (FTS) of hypercomplex neural networks (HCNNs) with mixed time-varying delays. Unlike finite-time synchronization (FNTS) based on initial conditions, the settling time of FTS can be adjusted to meet the needs. The state vector, weight matrices, activation functions, and input vectors of HCNNs are all hypercomplex numbers. The techniques used in complex-valued neural networks (CVNNs) and quaternion-valued neural networks (QVNNs) cannot be used directly with HCNNs because they do not work with eight or more dimensions. To begin with, the decomposition method is used to split the HCNNs into ((n+1)) real-valued neural networks (RVNNs) applying distributive law to handle non-commutativity and non-associativity. A nonlinear controller is constructed to synchronize the master-response systems of the HCNNs. Lyapunov-based method is used to prove the stability of an error system. The FTS of mixed time-varying delayed HCNNs is achieved using a suitable lemma, Lipschitz condition, appropriate Lyapunov functional construction, and designing suitable controllers. Two different algebraic criteria for settling time have been achieved by employing two distinct lemmas. It is demonstrated that the settling time derived from Lemma 1 produces a more precise result than that obtained from Lemma 2. Three numerical examples for CVNNs, QVNNs, and octonions-valued neural networks (OVNNs) are provided to demonstrate the efficacy and effectiveness of the proposed theoretical results.

Machine Reading Comprehension (MRC) is a task that enables machines to mirror key cognitive processes involving reading, comprehending a text passage, and answering questions about it. There has been significant progress in this task for English in recent years, where recent systems not only surpassed human-level performance but also demonstrated advancements in emulating complex human cognitive processes. However, the development of Arabic MRC has not kept pace due to language challenges and the lack of large-scale, high-quality datasets. Existing datasets are either small, low quality or released as a part of large multilingual corpora. We present the Arabic Question Answering Dataset (ArQuaD), a large MRC dataset for the Arabic language. The dataset comprises 16,020 questions posed by language experts on passages extracted from Arabic Wikipedia articles, where the answer to each question is a text segment from the corresponding reading passage. Besides providing various dataset analyses, we fine-tuned several pre-trained language models to obtain benchmark results. Among the compared methods, AraBERTv0.2-large achieved the best performance with an exact match of 68.95% and an F1-score of 87.15%. However, the significantly higher performance observed in human evaluations (exact match of 86% and F1-score of 95.5%) suggests a significant margin of possible improvement in future research. We release the dataset publicly at https://github.com/RashaMObeidat/ArQuAD to encourage further development of language-aware MRC models for the Arabic language.

One of the most important areas in medical image analysis is segmentation, in which raw image data is partitioned into structured and meaningful regions to gain further insights. By using Deep Neural Networks (DNN), AI-based automated segmentation algorithms can potentially assist physicians with more effective imaging-based diagnoses. However, since it is difficult to acquire high-quality ground truths for medical images and DNN hyperparameters require significant manual tuning, the results by DNN-based medical models might be limited. A potential solution is to combine multiple DNN models using ensemble learning. We propose a two-layer ensemble of deep learning models in which the prediction of each training image pixel made by each model in the first layer is used as the augmented data of the training image for the second layer of the ensemble. The prediction of the second layer is then combined by using a weight-based scheme which is found by solving linear regression problems. To the best of our knowledge, our paper is the first work which proposes a two-layer ensemble of deep learning models with an augmented data technique in medical image segmentation. Experiments conducted on five different medical image datasets for diverse segmentation tasks show that proposed method achieves better results in terms of several performance metrics compared to some well-known benchmark algorithms. Our proposed two-layer ensemble of deep learning models for segmentation of medical images shows effectiveness compared to several benchmark algorithms. The research can be expanded in several directions like image classification.

The image captioning task aims to generate descriptions from the main content of images. Recently, the Transformer with a self-attention mechanism has been widely used for the image captioning task, where the attention mechanism helps the encoder to generate image region features, and guides caption output in the decoder. However, the vanilla decoder uses a simple conventional self-attention mechanism, resulting in captions with poor semantic information and incomplete sentence logic. In this paper, we propose a novel attention block, Multi-Keys attention block, that fully enhances the relevance between explicit and implicit semantic information. Technically, the Multi-Keys attention block first concatenates the key vector and the value vector and spreads it into both the explicit channel and the implicit channel. Then, the “related value” is generated with more semantic information by applying the element-wise multiplication to them. Moreover, to perfect the sentence logic, the reverse key vector with another information flow is residually connected to the final attention result. We also apply the Multi-Keys attention block into the sentence decoder in the transformer named as Multi-Keys Transformer (MKTrans). The experiments demonstrate that our MKTrans achieves 138.6% CIDEr score on MS COCO “Karpathy” offline test split. The proposed Multi-Keys attention block and MKTrans model are proven to be more effective and superior than the state-of-the-art methods.

Deformable Medical Image Registration (DMIR) aims to establish precise anatomical alignment of multiple medical images. However, the existing U-shape networks encounter difficulties in efficiently transferring multi-scale feature information from the encoder to the decoder. To address this issue, we propose a novel backbone network called MFCTrans, which constructs effective feature connection in DMIR. Drawing inspiration from the attention mechanism observed in the human cognitive system, our proposed method employs a Feature Fusion and Assignment Transformer (FFAT) module and a Spatial Cross Attention Fusion (SCAF) module. The former facilitates the fusion of multi-channel features, while the latter guides the integration of multi-scale information. A Multiple Residual (MR) branch is also deployed between the encoder and FFAT to improve the network’s generalization. We conduct extensive qualitative and quantitative evaluations on the OASIS and LPBA40 datasets. The proposed method achieves higher Dice scores than Transmorph by 1.3% and 2.0% on the respective datasets while maintaining a comparable voxel folding percentage. Ablation studies analyze the impacts and efficiency of each component in the proposed method. In summary, our proposed network offers a promising framework for achieving high-quality medical image registration and holds significant potential for applications in computer vision and cognitive computation.

The field of cognitive computing, conversational AI has witnessed remarkable progress, largely driven by the development of the Generative Pre-trained Transformer (GPT) series, notably ChatGPT. These transformer-based models have revolutionized natural language understanding by effectively capturing context and long-range dependencies. In light of this, this paper conducts a comprehensive exploration of ChatGPT, encompassing its architectural design, training methodology, real-world applications, and future potential within the conversational AI landscape. The paper studies the ChatGPT ability for advanced control and responsiveness, exhibiting a superior capacity for comprehending language and generating precise, informative responses. The comprehensive survey depicts ChatGPT excels in sustaining context and engaging in multi-turn dialogues, thereby fostering more interactive and meaningful conversations. Furthermore, its adaptability for integration into various systems and scalability has broadened its applicability across diverse domains, including customer service, education, content generation, healthcare, gaming, research, and exploration. Additionally, the paper presents alternative conversational AI models, such as Amazon Codewhisperer, Google Bard (LaMDA), Microsoft Bing AI, DeepMind Sparrow, and Character AI, providing a comparative analysis that underscores ChatGPT’s advantages in terms of inference capabilities and future promise. Recognizing the evolution and profound impact of ChatGPT holds paramount significance for researchers and developers at the forefront of AI innovation. In a rapidly evolving conversational AI landscape, ChatGPT emerges as a pivotal player, capable of reshaping the way we interact with AI systems across a wide array of applications.

The utilization of word embeddings—powerful models computed through Neural Network architectures that encode words as vectors—has witnessed rapid growth across various Natural Language Processing applications, encompassing semantic analysis, information retrieval, dependency parsing, question answering, and machine translation. The efficacy of these tasks is strictly linked to the quality of the embeddings, underscoring the critical importance of evaluating and selecting optimal embedding models. While established procedures and benchmarks exist for intrinsic evaluation, the authors note a conspicuous absence of comprehensive evaluations of intrinsic embedding quality across multiple tasks. This paper introduces vec2best, a unified tool encompassing state-of-the-art intrinsic evaluation tasks across diverse benchmarks. vec2best furnishes the user with an extensive evaluation of word embedding models. It represents a framework for evaluating word embeddings trained using various methods and hyper-parameters on a range of tasks from the literature. The tool yields a holistic evaluation metric for each model called the PCE (Principal Component Evaluation). We conducted evaluations on 135 word embedding models, trained using GloVe, fastText, and word2vec, across four tasks integrated into vec2best (similarity, analogy, categorization, and outlier detection), along with their respective benchmarks. Additionally, we leveraged vec2best to optimize embedding hyper-parameter configurations in a real-world scenario. vec2best is conveniently accessible as a pip-installable Python package.

In today’s advanced applications, such as memory interfaces, feature-based detection, and sensory games, human-computer interaction (HCI) plays a pivotal role. A medical decision support system (MDSS) emerges from the integration of a data system with resources for medical decision-making. Within MDSS, human-computer interaction and perceptual medical decision-making stand out as two highly valuable technologies. Systems enabled by the Internet of Things (IoT), which leverage decentralized, diverse communication and networking technology to cater to a wide range of end-users, are referred to as pervasive computing. A challenging aspect of pervasive computing is ensuring transparency in interaction, managing administration levels, and accommodating varying tolerance levels for widely dispersed users. This paper presents a uniquely flexible MDSS framework designed to enhance end-user confidence in the availability of MDSS through ubiquitous IoT devices within the context of HCI. This architecture utilizes recurring training to assess resource allocation based on demand and collaborative characteristics. Projected resource requirements enable pervasive computing to better serve end-users by reducing latency and increasing communication speeds for MDSS in HCI. The primary goal of this framework is to simplify the management of terminal transitions by facilitating the allocation and utilization of resources for data transfer from peripheral technology. Experimental analysis is employed to estimate the framework’s performance, utilizing various metrics to demonstrate its consistency. These metrics encompass responsiveness, transaction success rates, processed demands, application caseloads, capacity utilization, and memory usage. The uniquely flexible and distributed computing framework optimizes request handling, network accuracy, and memory utilization, resulting in reduced transaction failures and lower latency, ultimately leading to shorter response times. The proposed UFDSS maintains a transaction failure rate below 25% with increasing requests and achieves 100 MHz bandwidth utilization, surpassing other techniques capped at 80 MHz. UFDSS exhibits a lower average latency of around 30 ms for a range of energy data inputs. This uniquely flexible MDSS framework showcases its potential to enhance MDSS availability through IoT devices within HCI contexts. By optimizing resource allocation and utilization, it successfully reduces latency, improves communication speeds, and ultimately leads to shorter response times, contributing to more efficient and reliable medical decision support. Further, integrating generative AI into MDSS for IoT-based HCI could also enhance data-driven decision support.

Deep Learning (DL) stands out as a leading model for processing high-dimensional data, where the nonlinear transformation of hidden layers effectively extracts features. However, these unexplainable features make DL a low interpretability model. Conversely, Bayesian network (BN) is transparent and highly interpretable, and it can be helpful for interpreting DL. To improve the interpretability of DL from the perspective of feature cognition, we propose the feature analysis network (FAN), a DL structure fused with BN. FAN retains the DL feature extraction capability and applies BN as the output layer to learn the relationships between the features and the outputs. These relationships can be probabilistically represented by the structure and parameters of the BN, intuitively. In a further study, a correlation clustering-based feature analysis network (cc-FAN) is proposed to detect the correlations among inputs and to preserve this information to explain the features’ physical meaning to a certain extent. To quantitatively evaluate the interpretability of the model, we design the network simplification and interpretability indicators separately. Experiments on eight datasets show that FAN has better interpretability than that of the other models with basically unchanged model accuracy and similar model complexities. On the radar effect mechanism dataset, from the feature structure-based relevance interpretability indicator, FAN is up to 4.8 times better than that of the other models, and cc-FAN is up to 21.5 times better than that of the other models. FAN and cc-FAN enhance the interpretability of the DL model structure from the aspects of features; moreover, based on the input correlations, cc-FAN can help us to better understand the physical meaning of features.

Introduction: Pupil size is a well-known indicator of low-level decision-making processes. However, it is unclear whether these involuntary eye data can represent information about the interwoven processes of hierarchical decision-making. In hierarchical decisions, high-level decision-making depends on the process of making low-level decisions, and the result of these interwoven processes is determined by feedback. Therefore, the exact cause of negative feedback is unclear, as it may be the result of low-level, high-level, or both low- and high-level incorrect decisions. In this study, we investigated the characteristics of eye data (pupil diameter) in the interwoven processes of hierarchical decision-making. Methods: We designed a hierarchical psychophysical experiment in which participants were asked to report their low- and high-level decisions and their confidence simultaneously on one of the colored bars. Participants received correct feedback in a trial when reporting both decisions correctly. During the experiment, the eye data of the participants were recorded by an eye-tracking device. Results: Our findings suggest that pupil size conveys information about high-level decisions as well. Furthermore, this study shows that three parameters (introduced in previous studies), negative feedback in successive trials, stimulus strength (uniformity with confidence), and decision urgency, are all represented in pupil size. Conclusion: The findings support the idea that involuntary eye data are influenced by decision-making-related brain activity in decision-making processes and not just visual stimulus features.