Decentralized tasks use machine learning models with certain assumptions. The first is the sharing of weights and feature extractors. The second is maintaining the privacy of the data. The idea of learning using multiple models can also be applied in parallel training, where a given model is trained on a different thread. This has applications in creating models based on federated learning, the Internet of Things and Digital Twins. This paper proposes a new neural network model that uses the atrous technique and attention modules. In addition, we propose a hierarchical-based training strategy, where the best model shares weights and is omitted during further training. This reduces the number of training epochs and increases the model's adaptability to a given set. The tests conducted on a publicly available medical database indicate high learning potential for both the proposed model and the hierarchical learning strategy.

Agricultural health monitoring is a critical task in ensuring the stability of modern agriculture. Many plant diseases share visual similarities, making manual inspection both time consuming and error prone, which is why robust and adaptable disease detection frameworks are not only desirable but essential to maintaining a resilient agricultural ecosystem. In this paper, we propose a hybrid federated learning (FL) framework that integrates a globally shared feature extractor with a client-specific self-attentive branch and classifier. The proposed framework uses a global model with both globally shared and client-tailored branches to achieve better performance for specialized tasks in decentralized training scenarios. The experiments were carried out on a Plant Village data set in a scenario, where each client represented a different crop type and faced a different leaf disease classification problem. The proposed solution revolved around the clients sharing the global weights, thus simultaneously contributing towards better feature extraction of the common leaf features, while the specialized segment of the model focused on proper interpretation of the extracted features (via cross-attention mechanism) and direct classification. The results obtained demonstrate the effectiveness of the proposed approach over standard local training, as training with the proposed hybrid FL framework resulted in a perfect classification of the precision 100% of apple leaf disease.

In the realm of Eulerian-based fluid dynamics simulations, the computational demands associated with the number of the projection steps, particularly the management of the extensive linear system stemming from the Gauss-Seidel equation, pose a significant temporal and computational burden. In this research paper, we propose a custom Deep Learning (DL) approach to the projection method leveraging machine learning paradigms, specifically integrating novel Deep Neural Networks (DNN) into a custom made 2D smoke simulator. Proposed solution offers a high quality output, comparable with naive techniques, while using less computational memory and fraction of solving time. The effectiveness of our proposed approach has been validated through testing across a spectrum of smoke scenes, chosen to substantially differ from the training dataset. The demonstrated outcomes underscore the accelerated computational performance and extrapolation capabilities inherent in our method, thus highlighting its adaptability and robustness in addressing simulation scenarios beyond the training dataset parameters.

Abstract The current proofreading algorithms for action regulation mainly recover the 3D structure and action information of non-rigid objects from image sequences by factorization. Most of algorithms assume that the camera model is an affine model. This assumption only holds if the size and depth of the object change very little relative to the distance from the object to the camera, which is in the case of fixed-shape basis. When the object is very close to the camera, this assumption causes a large reconstruction error. This paper solves this problem by the intelligent proofreading algorithms for remote skiing teaching actions based on variable shape basis. Firstly, the improved Retinex algorithm is used to enhance the multi-frame video images of skiing actions to make the action details more prominent. Then, measurement matrix is calculated after eliminating the translation vector by coordinate transformation. Under the condition of rank constraint, the measurement matrix is decomposed by singular value decomposition algorithm, and the correct shape basis structure of 3D action features can be obtained by using the variable shape basis. Finally, by randomly initializing a parameter, the optimized parameter and the least square algorithm are used to optimize the randomly initialized parameter further. The iteration until the convergence of the objective function can be used to calculate the deformation degree of the actions. The test results show that this algorithm improves the proofreading accuracy of action regulation in skiing teaching, and the proofreading results of various uploaded sliding actions are correct, which can be applied to remote skiing teaching and community learning.