The Gradient Boosting Decision Tree (GBDT) is a popular machine learning model for various tasks in recent years. In this paper, we study how to improve model accuracy of GBDT while preserving the strong guarantee of differential privacy. Sensitivity and privacy budget are two key design aspects for the effectiveness of differential private models. Existing solutions for GBDT with differential privacy suffer from the significant accuracy loss due to too loose sensitivity bounds and ineffective privacy budget allocations (especially across different trees in the GBDT model). Loose sensitivity bounds lead to more noise to obtain a fixed privacy level. Ineffective privacy budget allocations worsen the accuracy loss especially when the number of trees is large. Therefore, we propose a new GBDT training algorithm that achieves tighter sensitivity bounds and more effective noise allocations. Specifically, by investigating the property of gradient and the contribution of each tree in GBDTs, we propose to adaptively control the gradients of training data for each iteration and leaf node clipping in order to tighten the sensitivity bounds. Furthermore, we design a novel boosting framework to allocate the privacy budget between trees so that the accuracy loss can be further reduced. Our experiments show that our approach can achieve much better model accuracy than other baselines.
Flexible full-textile pressure sensor is able to integrate with clothing directly, which has drawn extensive attention from scholars recently. But the realization of flexible full-textile pressure sensor with high sensitivity, wide detection range, and long working life remains challenge. Complex recognition tasks necessitate intricate sensor arrays that require extensive data processing and are susceptible to damage. The human skin is capable of interpreting tactile signals, such as sliding, by encoding pressure changes and performing complex perceptual tasks. Inspired by the skin, we have developed a simple dip-and-dry approach to fabricate a full-textile pressure sensor with signal transmission layers, protective layers, and sensing layers. The sensor achieves high sensitivity (2.16 kPa
−1
), ultrawide detection range (0 to 155.485 kPa), impressive mechanical stability of 1 million loading/unloading cycles without fatigue, and low material cost. The signal transmission layers that collect local signals enable real-world complicated task recognition through one single sensor. We developed an artificial Internet of Things system utilizing a single sensor, which successfully achieved high accuracy in 4 tasks, including handwriting digit recognition and human activity recognition. The results demonstrate that skin-inspired full-textile sensor paves a promising route toward the development of electronic textiles with important potential in real-world applications, including human–machine interaction and human activity detection.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.