data load when using transmission sockets. At the same time, it reduces the dimensionality of the extracted data for the client-side transfer learning combinator, which has various purposes; ne-tuning model enhances personal prediction for dierent purposes and personalizes user data when merging EDCN and Decoder to form AutoEncoder (AE) variant. us, the solution helps to reproduce new data with good characteristics associated with shared data (baseline), partially solving the non-IID problem, and achieving reinforced learning and repetition.c. Research outcomee excellent performance demonstrated by Fed xData in smartphone-based human activity recognition. e test results show that Fed xData signicantly improves recogni-tion accuracy compared to general and clearing models. Fed xData is extensible and can be the standard framework for many healthcare applications. Specically, the Fed’s xData framework is designed and applied to recognize falls and strokes in the elderly. User privacy is well protected and achieves good performance in real-life scenarios. is research can widely monitor health status and warn of falls and strokes in a narrow range such as houses, apartments, hospitals, and nursing homes.3.2.2FedMCRNN: Federated learning using Multiple Convolutional Recurrent Neural Networks for sleep quality predictiona. Sleep quality prediction from IoT environmental data Sleep plays a vital role in helping the body rest, restore and regenerate energy for the operation of organs in the body, especially the brain. e circadian clock regulates the human sleep-wake cycle in the brain, constantly balancing the sleep and wake times of the human body. Good sleep is a fundamental part of a healthy life, which improves all bodily functions and mental state. Using the data collected in daily activities such as IoT, such as a smartwatch, [14] introduced an FL application solution to predict sleep quality.b. MethodologyIn this study, the rst important issue is data processing since daily human activities include many dierent activi-ties. So, what actions aect the quality of sleep is a top concern. We have researched and found that in this data-set, only six important activity categories were used to assess sleep quality eectively (calories, distance, sedentary minutes, lightly activity minutes, active minutes, and very active minutes). ose categories are used to predict sleep-related outputs, and the outcome is a per-sleep breakdown of the rest into periods of light, deep, REM sleeps, and time awake, where sleep eciency is the most concerning part. Next, we introduce how to choose the future prediction by the data windowing method. e purpose is that we can predict outcomes by adding data appropriately. And -nally, In the same direction on Edge AI federated learning with IoT devices, we also proposed operating federated multiple convolutional neural networks (FedMCRNN) (Fig. 8) to predict sleep quality. e advantage of this model is to benet from traditional CRNN to develop a new neural network model.c. Research outcomeIn experimental results, we measure the performance of the FedMCRNN in many-to-one and many-to-many cases using a variety of metrics and compare it with tradi-tional machine learning models. e results show that FedMCRNN predicts quality intention reliably, with 96.774% and 68.721% accuracies for the two cases, many-to-one and many-to-many, respectively. Besides, other metrics have better value than methods. e results also show that FedMCRNN performs better than previous most advanced methods for predicting sleep quality and clearly shows which features inuence sleep quality. Our ndings have implications for developing AI doctors. ig. F8 Federated Multiple Convolutional Recurrent Neural Networks for Sleep Quality Prediction1834-4 パブリック・プライベートデータを活用した予測モデリングを実現する連合学習によるエッジAI
元のページ ../index.html#189