Abstract: A multi-layer federated learning method based on distributed clustering is provided, which comprises the following steps. Computing a feature distribution similarity for each of participating nodes with non-(non-independent and identically distributed) data sets, and grouping these nodes into plural clusters by the feature distribution similarity. Updating local model of nodes of each cluster by a federated learning algorithm, and inputting these nodes into a multi-layer aggregation mechanism. Terminating the operation of the multi-layer aggregation mechanism until the clustering result meets a required demand. Furthermore, we implement a blockchain-based multi-layer federated learning system, including model aggregation module, API module, time-series synchronization module, and IPFS, based on distributed clustering architecture. The learning performance is proven to effectively improved.

Abstract: A medical data ownership management method comprises: Granting a first authority to a medical unit where the first authority allows the medical unit to generate a first medical object with its corresponding first non-fungible token (NFT) storing on a blockchain. Granting a second authority to a research unit, where the second authority allows the research unit to access the first medical object, and the research unit processes the first medical object to generate the second medical object with its corresponding second NFT storing on the blockchain. Finally, granting a third authority to a user unit, where the third authority allows the user unit to access the second medical object. The access control and ownership transfer of each medial object in the blockchain are achieved by trading the corresponding NFT and modifying its subordinate list. With the management method, the security and the privacy of the medical data will be enhanced.

Abstract: An image authentication method and a real-time product authentication system are provided. The method includes the following steps: inputting a training image in a training phase; segmenting the training image into a plurality of training sub-images; and training the deep learning model by using the plurality of training sub-images, wherein each of the plurality of training images includes a global feature property. The steps in the testing phase include: inputting a testing image and segmenting the testing image into a plurality of testing sub-images; performing a test on each of the plurality of testing sub-images and checking whether each of the plurality of testing sub-images is associated with one of the plurality of categories; and outputting an authentication result to identify whether the testing image is associated with one of the plurality of categories.

Abstract: An intelligent gait analyzing apparatus comprises a first camera and an electronic device. The first camera captures a first video in which the patient walks and a second video in which the patient changes a posture from standing up to sitting down. A second camera may be used to capture a three-dimensional point cloud diagram of the patient to generate a chest skeleton and an abdomen skeleton. A 3D skeleton model and a moving diagram may be created based on the first video. The memory stores a long short-term memory model, a transformer machine learning model and an algorithm of modified gravity center. The graphic processor calculates a gravity center deviation based on the moving diagram, the chest skeleton and the abdomen skeleton. The graphic processor acquires a stride, a pace, a humpbacked value and a risk of falling.

Abstract: An image authentication method and a real-time product authentication system are provided. The method includes the following steps: inputting a training image in a training phase; segmenting the training image into a plurality of training sub-images; and training the deep learning model by using the plurality of training sub-images, wherein each of the plurality of training images includes a global feature property. The steps in the testing phase include: inputting a testing image and segmenting the testing image into a plurality of testing sub-images; performing a test on each of the plurality of testing sub-images and checking whether each of the plurality of testing sub-images is associated with one of the plurality of categories; and outputting an authentication result to identify whether the testing image is associated with one of the plurality of categories.