Search

Data-free knowledge distillation is a challenging model lightweight task for scenarios in which the original dataset is not available. Previous methods require a lot of extra computational costs to update one or more generators and their naive imitate-learning lead to lower distillation efficiency. Based on these observations, we first propose an efficient unlabeled sample selection method to replace high computational generators and focus on improving the training efficiency of the selected samples. Then, a class-dropping mechanism is designed to suppress the label noise caused by the data domain shifts. Finally, we propose a distillation method that incorporates explicit features and implicit structured relations to improve the effect of distillation. Experimental results show that our method can quickly converge and obtain higher accuracy than other state-of-the-art methods., Comment: accepted for ICASSP 2023

Large-scale deep learning models have achieved great performance based on large-scale datasets. Moreover, the existing Adversarial Training (AT) can further improve the robustness of these large models. However, these large models are difficult to deploy to mobile devices, and the effect of AT on small models is very limited. In addition, the data privacy issue (e.g., face data and diagnosis report) may lead to the original data being unavailable, which relies on data-free knowledge distillation technology for training. To tackle these issues, we propose a challenging novel task called Data-Free Adversarial Robustness Distillation (DFARD), which tries to train small, easily deployable, robust models without relying on the original data. We find the combination of existing techniques resulted in degraded model performance due to fixed training objectives and scarce information content. First, an interactive strategy is designed for more efficient knowledge transfer to find more suitable training objectives at each epoch. Then, we explore an adaptive balance method to suppress information loss and obtain more data information than previous methods. Experiments show that our method improves baseline performance on the novel task.

Context-Aware Emotion Recognition (CAER) is a crucial and challenging task that aims to perceive the emotional states of the target person with contextual information. Recent approaches invariably focus on designing sophisticated architectures or mechanisms to extract seemingly meaningful representations from subjects and contexts. However, a long-overlooked issue is that a context bias in existing datasets leads to a significantly unbalanced distribution of emotional states among different context scenarios. Concretely, the harmful bias is a confounder that misleads existing models to learn spurious correlations based on conventional likelihood estimation, significantly limiting the models' performance. To tackle the issue, this paper provides a causality-based perspective to disentangle the models from the impact of such bias, and formulate the causalities among variables in the CAER task via a tailored causal graph. Then, we propose a Contextual Causal Intervention Module (CCIM) based on the backdoor adjustment to de-confound the confounder and exploit the true causal effect for model training. CCIM is plug-in and model-agnostic, which improves diverse state-of-the-art approaches by considerable margins. Extensive experiments on three benchmark datasets demonstrate the effectiveness of our CCIM and the significance of causal insight., Comment: Accepted by CVPR 2023. CCIM is available at https://github.com/ydk122024/CCIM

Through the research of CPCI hot plug and avionics data bus ARINC429, this paper analyzes and proposes the improvement direction needed to support the high availability hot plug for ARINC429 in CPCI. In addition, this paper also designed the high availability hot plug ARINC429-CPCI board in practice, and provided the necessary support from hardware and software to achieve the target functions. This paper also summarizes the relevant theory and design experience for the application of high availability hot plug technology in aeronautical measurement and control systems.

OBJECTIVES: Previous studies have shown that brief mindfulness trainings can have significant analgesic effects. However, the effects of the various components of mindfulness on pain analgesia are not well understood. The objective of this study was to examine the effects of two components of mindfulness interventions - attention and acceptance on pain analgesia. METHODS: One hundred and nineteen healthy college students without prior mindfulness experience underwent a cold pressor test to measure pain tolerance before and after the training. Pain intensity, tolerance, distress, threshold and endurance time were also tested. Participants were randomly assigned to one of four conditions: (1) acceptance of pain, (2) attention to pain, (3) acceptance of and attention to pain, or (4) control. RESULTS: The results showed that both the acceptance strategy and the combined acceptance and attention group increased pain endurance and tolerance after training. Furthermore, acceptance group had longer pain endurance time and tolerance time than attention group and control group. CONCLUSIONS: These results suggest that acceptance of pain is more important than attention to pain. Study limitations and future research directions are discussed.

Immune cells in tumor microenvironment play a prominent role in tumor progression and metastasis. MicroRNA-155 (miR-155) represents an important player in innate and adaptive immunity by regulating differentiation, maturation and activation of macrophages, dendritic cells, B cells and T cells. However, the role of miR-155 expression in immune cells in solid tumor development is less elucidated. Our current study showed that both B16-F10 melanoma and Lewis lung carcinoma tumors grew much faster in bic/miR-155 knockout (miR-155(-/-) ) mice along with an increase of myeloid-derived suppressor cells (MDSCs) accumulation in tumors, compared to that in wild-type mice. Bone marrow transplantation study showed that bone marrow miR-155 deficiency could replicate the above tumor-promoting phenotype. In vitro study demonstrated that tumor-infiltrating miR-155(-/-) MDSCs showed greater migration ability and expressed higher level of multiple chemokines. Furthermore, we found that the level of HIF-1α, a direct target of miR-155, was increased in miR-155 deficient MDSCs, and that the increased HIF-1α upregulated CXCL1, CXCL3 and CXCL8 expression in MDSCs, contributing to the enhanced recruitment of miR-155(-/-) MDSCs to the tumors. Moreover, miR-155(-/-) MDSCs showed enhanced immunosuppressive and pro-angiogenic capacities. Taken together, our study, for the first time, demonstrated that miR-155 deficiency promoted solid tumor growth through increasing the recruitment of MDSCs to tumor microenvironment and enhancing the tumor-promoting functions of the recruited MDSCs. Thus, upregulating miR-155 expression in MDSCs may be developed as a therapeutic approach to halt tumor development.