CAAI Transactions on Intelligence Technology最新文献

The prevalence of digestive system tumours (DST) poses a significant challenge in the global crusade against cancer. These neoplasms constitute 20% of all documented cancer diagnoses and contribute to 22.5% of cancer-related fatalities. The accurate diagnosis of DST is paramount for vigilant patient monitoring and the judicious selection of optimal treatments. Addressing this challenge, the authors introduce a novel methodology, denominated as the Multi-omics Graph Transformer Convolutional Network (MGTCN). This innovative approach aims to discern various DST tumour types and proficiently discern between early-late stage tumours, ensuring a high degree of accuracy. The MGTCN model incorporates the Graph Transformer Layer framework to meticulously transform the multi-omics adjacency matrix, thereby illuminating potential associations among diverse samples. A rigorous experimental evaluation was undertaken on the DST dataset from The Cancer Genome Atlas to scrutinise the efficacy of the MGTCN model. The outcomes unequivocally underscore the efficiency and precision of MGTCN in diagnosing diverse DST tumour types and successfully discriminating between early-late stage DST cases. The source code for this groundbreaking study is readily accessible for download at https://github.com/bigone1/MGTCN.

Virtual reality (VR) technology revitalises rehabilitation training by creating rich, interactive virtual rehabilitation scenes and tasks that deeply engage patients. Robotics with immersive VR environments have the potential to significantly enhance the sense of immersion for patients during training. This paper proposes a rehabilitation robot system. The system integrates a VR environment, the exoskeleton entity, and research on rehabilitation assessment metrics derived from surface electromyographic signal (sEMG). Employing more realistic and engaging virtual stimuli, this method guides patients to actively participate, thereby enhancing the effectiveness of neural connection reconstruction—an essential aspect of rehabilitation. Furthermore, this study introduces a muscle activation model that merges linear and non-linear states of muscle, avoiding the impact of non-linear shape factors on model accuracy present in traditional models. A muscle strength assessment model based on optimised generalised regression (WOA-GRNN) is also proposed, with a root mean square error of 0.017,347 and a mean absolute percentage error of 1.2461%, serving as critical assessment indicators for the effectiveness of rehabilitation. Finally, the system is preliminarily applied in human movement experiments, validating the practicality and potential effectiveness of VR-centred rehabilitation strategies in medical recovery.

The Runge–Kutta optimiser (RUN) algorithm, renowned for its powerful optimisation capabilities, faces challenges in dealing with increasing complexity in real-world problems. Specifically, it shows deficiencies in terms of limited local exploration capabilities and less precise solutions. Therefore, this research aims to integrate the topological search (TS) mechanism with the gradient search rule (GSR) into the framework of RUN, introducing an enhanced algorithm called TGRUN to improve the performance of the original algorithm. The TS mechanism employs a circular topological scheme to conduct a thorough exploration of solution regions surrounding each solution, enabling a careful examination of valuable solution areas and enhancing the algorithm’s effectiveness in local exploration. To prevent the algorithm from becoming trapped in local optima, the GSR also integrates gradient descent principles to direct the algorithm in a wider investigation of the global solution space. This study conducted a serious of experiments on the IEEE CEC2017 comprehensive benchmark function to assess the enhanced effectiveness of TGRUN. Additionally, the evaluation includes real-world engineering design and feature selection problems serving as an additional test for assessing the optimisation capabilities of the algorithm. The validation outcomes indicate a significant improvement in the optimisation capabilities and solution accuracy of TGRUN.

In this study, a Grey-box (GB) model was developed to predict the optimum mass flow rates of inlet streams of a Shell and Tube Heat Exchanger (STHE) under varying process conditions. Aspen Exchanger Design and Rating (Aspen-EDR) was initially used to construct a first principle model (FP) of the STHE using industrial data. The Genetic Algorithm (GA) was incorporated into the FP model to attain the minimum exit temperature for the hot kerosene process stream under varying process conditions. A dataset comprised of optimum process conditions was generated through FP-GA integration and was utilised to develop an Artificial Neural Networks (ANN) model. Subsequently, the ANN model was merged with the FP model by substituting the GA, to form a GB model. The developed GB model, that is, ANN and FP integration, achieved higher effectiveness and lower outlet temperature than those derived through the standalone FP model. Performance of the GB framework was also comparable to the FP-GA approach but it significantly reduced the computation time required for estimating the optimum process conditions. The proposed GB-based method improved the STHE's ability to extract energy from the process stream and strengthened its resilience to cope with diverse process conditions.

Real-world data always exhibit an imbalanced and long-tailed distribution, which leads to poor performance for neural network-based classification. Existing methods mainly tackle this problem by reweighting the loss function or rebalancing the classifier. However, one crucial aspect overlooked by previous research studies is the imbalanced feature space problem caused by the imbalanced angle distribution. In this paper, the authors shed light on the significance of the angle distribution in achieving a balanced feature space, which is essential for improving model performance under long-tailed distributions. Nevertheless, it is challenging to effectively balance both the classifier norms and angle distribution due to problems such as the low feature norm. To tackle these challenges, the authors first thoroughly analyse the classifier and feature space by decoupling the classification logits into three key components: classifier norm (i.e. the magnitude of the classifier vector), feature norm (i.e. the magnitude of the feature vector), and cosine similarity between the classifier vector and feature vector. In this way, the authors analyse the change of each component in the training process and reveal three critical problems that should be solved, that is, the imbalanced angle distribution, the lack of feature discrimination, and the low feature norm. Drawing from this analysis, the authors propose a novel loss function that incorporates hyperspherical uniformity, additive angular margin, and feature norm regularisation. Each component of the loss function addresses a specific problem and synergistically contributes to achieving a balanced classifier and feature space. The authors conduct extensive experiments on three popular benchmark datasets including CIFAR-10/100-LT, ImageNet-LT, and iNaturalist 2018. The experimental results demonstrate that the authors’ loss function outperforms several previous state-of-the-art methods in addressing the challenges posed by imbalanced and long-tailed datasets, that is, by improving upon the best-performing baselines on CIFAR-100-LT by 1.34, 1.41, 1.41 and 1.33, respectively.

A mixed adaptive dynamic programming (ADP) scheme based on zero-sum game theory is developed to address optimal control problems of autonomous underwater vehicle (AUV) systems subject to disturbances and safe constraints. By combining prior dynamic knowledge and actual sampled data, the proposed approach effectively mitigates the defect caused by the inaccurate dynamic model and significantly improves the training speed of the ADP algorithm. Initially, the dataset is enriched with sufficient reference data collected based on a nominal model without considering modelling bias. Also, the control object interacts with the real environment and continuously gathers adequate sampled data in the dataset. To comprehensively leverage the advantages of model-based and model-free methods during training, an adaptive tuning factor is introduced based on the dataset that possesses model-referenced information and conforms to the distribution of the real-world environment, which balances the influence of model-based control law and data-driven policy gradient on the direction of policy improvement. As a result, the proposed approach accelerates the learning speed compared to data-driven methods, concurrently also enhancing the tracking performance in comparison to model-based control methods. Moreover, the optimal control problem under disturbances is formulated as a zero-sum game, and the actor-critic-disturbance framework is introduced to approximate the optimal control input, cost function, and disturbance policy, respectively. Furthermore, the convergence property of the proposed algorithm based on the value iteration method is analysed. Finally, an example of AUV path following based on the improved line-of-sight guidance is presented to demonstrate the effectiveness of the proposed method.

Efficiently executing inference tasks of deep neural networks on devices with limited resources poses a significant load in IoT systems. To alleviate the load, one innovative method is branching that adds extra layers with classification exits to a pre-trained model, enabling inputs with high-confidence predictions to exit early, thus reducing inference cost. However, branching networks, not originally tailored for IoT environments, are susceptible to noisy and out-of-distribution (OOD) data, and they demand additional training for optimal performance. The authors introduce BrevisNet, a novel branching methodology designed for creating on-device branching models that are both resource-adaptive and noise-robust for IoT applications. The method leverages the refined uncertainty estimation capabilities of Dirichlet distributions for classification predictions, combined with the superior OOD detection of energy-based models. The authors propose a unique training approach and thresholding technique that enhances the precision of branch predictions, offering robustness against noise and OOD inputs. The findings demonstrate that BrevisNet surpasses existing branching techniques in training efficiency, accuracy, overall performance, and robustness.

Offline policy evaluation, evaluating and selecting complex policies for decision-making by only using offline datasets is important in reinforcement learning. At present, the model-based offline policy evaluation (MBOPE) is widely welcomed because of its easy to implement and good performance. MBOPE directly approximates the unknown value of a given policy using the Monte Carlo method given the estimated transition and reward functions of the environment. Usually, multiple models are trained, and then one of them is selected to be used. However, a challenge remains in selecting an appropriate model from those trained for further use. The authors first analyse the upper bound of the difference between the approximated value and the unknown true value. Theoretical results show that this difference is related to the trajectories generated by the given policy on the learnt model and the prediction error of the transition and reward functions at these generated data points. Based on the theoretical results, a new criterion is proposed to tell which trained model is better suited for evaluating the given policy. At last, the effectiveness of the proposed criterion is demonstrated on both benchmark and synthetic offline datasets.