Partial multi-label learning (PML) aims to train a classifier on dataset whose instances are over-annotated with not only relevant labels but also irrelevant labels, which is common when datasets are collected from crowd-sourcing platform. Existing works primarily approach it from a curriculum learning perspective, leveraging the memorization effect to disambiguate noisy labels and produce robust predictions. However, these methods are based on non-adaptive weighting functions and lack theoretical guidance for optimal weighting. To overcome these issues, a calibrated disambiguation model named PML-CD is proposed. We firstly formulate the optimal weighting function for curriculum-based disambiguation, which is equivalent to the calibration of the model's predicted confidences, thus provide a guidance for curriculum designing. To obtain the optimal weighting function from PML dataset during the training, a transferable calibrator is designed, which takes the histogram of positive samples' confidences as input, and outputs the optimal curriculum weighting for training. Prototype alignment regularization is also proposed to promote the model's performance. Experiments conducted on Pascal VOC, MS-COCO, NUS-WIDE and CUB have verified that our method outperforms existing state-of-the-art PML methods.

Predicting drug-disease associations (DDAs) is a crucial task in drug discovery and repositioning, as it enables the identification of novel therapeutic indications for existing drugs. However, current technologies for predicting these associations face limitations, such as challenges in data integration, difficulty in capturing multi-relational biological interactions, and limited generalizability. To address these issues, we present a network-based framework that integrates heterogeneous biological data-including drugs, proteins, biological functions, and diseases-into a unified graph structure. To effectively model the complex and multi-relational nature of this network, we employ a Relational Graph Convolutional Network (RGCN), which explicitly captures various types of biological interactions. Our model named DDA-RGCN achieves superior predictive performance, with an AUC of 0.9745 and an AUPR of 0.9728, outperforming several baseline and state-of-the-art methods. Case studies further demonstrate the model's potential in uncovering novel drug indications, supported by literature evidence. These findings underscore the value of combining heterogeneous biological networks with graph neural network techniques and highlight the potential of our RGCN-based approach to advance computational drug repositioning and support personalized medicine initiatives.

In the field of software engineering, the automated analysis of log data is crucial for operations and maintenance teams. This study introduces LLMDPP, a novel log parser that leverages Large Language Models (LLMs) and Determinantal Point Process (DPP) sampling techniques to enhance the efficiency and accuracy of online log parsing. LLMDPP transforms raw log messages into structured log templates through a few-shot learning, thus simplifying the processing and analysis of log data. The study explores the accuracy of LLMDPP in log-parsing tasks and compares the effectiveness of different encoding functions (TF-IDF and Flan-T5-small embedding layer) in DPP sampling. Experimental results show that LLMDPP outperforms traditional methods in both Global Accuracy and Parsing Accuracy, with the semantic information encoding function performing better when the number of samples is low. Furthermore, we simulate an online scenario to evaluate the parsing effectiveness of different sampling methods on unseen log datasets. The results indicate that the DPP sampling method has an advantage in maintaining sample diversity and fairness, which can improve parsing accuracy.

Jumping with controllable elevation is significant for insect-scale robots to improve terrain adaptability and to escape from risks. However, jumping robots based on soft materials with low stiffness cannot transmit displacement precisely, exhibiting poor control of jumping. Here, we propose a modified two-bars catapult mechanism combined with an asynchronous sequential releasing strategy to realize elevation controllable jumping. In this work, an 80 mg prototype robot, 56 (long) × 29 mm (wide) × 3 mm (high) mm in size, is designed with the controllable elevation range from 63° to 112°. The soft robot is mainly composed of a shape memory alloy actuator and four electrostatic pads acting as the lock/release structures. Elevation control is realized by asynchronously releasing the electrostatic pads in a small time interval (about 10 ms). A maximum jump height of 62 mm and a maximum half-distance of 41 mm are also achieved.