Crowd evacuation in different situations is an important topic in the research field of safety. This paper presents a hybrid model for heterogeneous pedestrian evacuation simulation. Our adaptive agent-based model (AB...Crowd evacuation in different situations is an important topic in the research field of safety. This paper presents a hybrid model for heterogeneous pedestrian evacuation simulation. Our adaptive agent-based model (ABM) combines the strength of human crowd behavior description from classical social force models with discrete dynamics expression from cellular automaton models by extending the conception of floor field. Several important factors which may influence the results of decision-making of pedestrians are taken into consideration, such as the location of sign, the attraction of exit, and the interaction among pedestrians. To compare the effect of information on the pedestrians, we construct three decision-making mechanisms with different assumptions. To validate these three simulation models, we compare the numerical results from different perspectives with rational range in the case study where the Tampere Theater evacuation was carried out. The ABM framework is open for rules modification and could be applied to different building plans and has implication for architectural design of gates and signs in order to increase the evacuation efficiency.展开更多
Federated Learning(FL) heralds a paradigm shift in the training of artificial intelligence(AI) models by fostering collaborative model training while safeguarding client data privacy. In sectors where data sensitivity...Federated Learning(FL) heralds a paradigm shift in the training of artificial intelligence(AI) models by fostering collaborative model training while safeguarding client data privacy. In sectors where data sensitivity and AI model security are of paramount importance, such as fintech and biomedicine, maintaining the utility of models without compromising privacy is crucial with the growing application of AI technologies. Therefore, the adoption of FL is attracting significant attention. However, traditional FL methods are susceptible to Deep Leakage from Gradients(DLG) attacks, and typical defensive strategies in current research, such as secure multi-party computation and diferential privacy, often lead to excessive computational costs or significant decreases in model accuracy. To address DLG attacks in FL, this study introduces VAEFL, an innovative FL framework that incorporates Variational Autoencoders(VAEs) to enhance privacy protection without undermining the predictive prowess of the models. VAEFL strategically partitions the model into a private encoder and a public decoder. The private encoder, remaining local, transmutes sensitive data into a latent space fortified for privacy, while the public decoder and classifier, through collaborative training across clients, learn to derive precise predictions from the encoded data. This bifurcation ensures that sensitive data attributes are not disclosed, circumventing gradient leakage attacks and simultaneously allowing the global model to benefit from the diverse knowledge of client datasets. Comprehensive experiments demonstrate that VAEFL not only surpasses standard FL benchmarks in privacy preservation but also maintains competitive performance in predictive tasks. VAEFL thus establishes a novel equilibrium between data privacy and model utility, ofering a secure and efficient FL approach for the sensitive application of FL in the financial domain.展开更多
基金the Natural Science Foundation of Shanghai (No. 18ZR1420200)the National Natural Science Foundation of China (No. 61603253)the China Postdoctoral Science Foundation Funded Project (No. 2016M601598)。
文摘Crowd evacuation in different situations is an important topic in the research field of safety. This paper presents a hybrid model for heterogeneous pedestrian evacuation simulation. Our adaptive agent-based model (ABM) combines the strength of human crowd behavior description from classical social force models with discrete dynamics expression from cellular automaton models by extending the conception of floor field. Several important factors which may influence the results of decision-making of pedestrians are taken into consideration, such as the location of sign, the attraction of exit, and the interaction among pedestrians. To compare the effect of information on the pedestrians, we construct three decision-making mechanisms with different assumptions. To validate these three simulation models, we compare the numerical results from different perspectives with rational range in the case study where the Tampere Theater evacuation was carried out. The ABM framework is open for rules modification and could be applied to different building plans and has implication for architectural design of gates and signs in order to increase the evacuation efficiency.
基金supported by the Yangtze River Delta Science and Technology Innovation Community Joint Research Project (2022CSJGG0800)the Shanghai Science and Technology Project (22510761000)
文摘Federated Learning(FL) heralds a paradigm shift in the training of artificial intelligence(AI) models by fostering collaborative model training while safeguarding client data privacy. In sectors where data sensitivity and AI model security are of paramount importance, such as fintech and biomedicine, maintaining the utility of models without compromising privacy is crucial with the growing application of AI technologies. Therefore, the adoption of FL is attracting significant attention. However, traditional FL methods are susceptible to Deep Leakage from Gradients(DLG) attacks, and typical defensive strategies in current research, such as secure multi-party computation and diferential privacy, often lead to excessive computational costs or significant decreases in model accuracy. To address DLG attacks in FL, this study introduces VAEFL, an innovative FL framework that incorporates Variational Autoencoders(VAEs) to enhance privacy protection without undermining the predictive prowess of the models. VAEFL strategically partitions the model into a private encoder and a public decoder. The private encoder, remaining local, transmutes sensitive data into a latent space fortified for privacy, while the public decoder and classifier, through collaborative training across clients, learn to derive precise predictions from the encoded data. This bifurcation ensures that sensitive data attributes are not disclosed, circumventing gradient leakage attacks and simultaneously allowing the global model to benefit from the diverse knowledge of client datasets. Comprehensive experiments demonstrate that VAEFL not only surpasses standard FL benchmarks in privacy preservation but also maintains competitive performance in predictive tasks. VAEFL thus establishes a novel equilibrium between data privacy and model utility, ofering a secure and efficient FL approach for the sensitive application of FL in the financial domain.
