Protect privacy of deep classification networks by exploiting their generative power

Machine Learning - Research showed that deep learning models are vulnerable to membership inference attacks, which aim to determine if an example is in the training set of the model. We propose a...     

CERTAIN CRITERION FOR THE FORMATION OF AMORPHOUS SOLIDS

<正> 根据成核理论,当仅考虑均匀核化时,在过冷熔体中结晶成核率可用下式表示: I=N_v~°kT/(3πα_°~3η)exp(-16πα~3β/3T_rΔT_r~2) (1)其中α=(N_v~°)~(1/3)σ/ΔH_m,β=1ΔH_m/RT_m=T/T_m, ΔT_r=T_m-T/T_m.各参量分别表示:I,成核率(cm~(-3)·s~(-1);N_v~°单位体积中的分子数(cm~(-3);η,粘滞系数(P);σ,比表面能(cal·cm~(-2));ΔH_m,熔化时的焓变(cal·mol~(-1));N,阿佛加德罗常数;v,克分子体积;T,熔体温度(K);T_m,熔点(K);k,玻尔兹曼常数;R,气体常数;α_0,分子直径。     

不同活化剂对石油焦基活性炭孔结构的影响

以石油焦为原料 ,Na OH,KOH和 Na2 CO3 为活化剂制备活性炭 ,采用氮气吸附考察了不同活化剂对活性炭的比表面积、中孔和微孔孔径分布、孔容积及平均孔径等孔结构的影响 .结果表明 :KOH活化制备的活性炭包含 1 nm的微孔和 4nm的中孔 ,总孔容 0 .648cm3 /g,比表面积大 ;Na OH制备的活性炭以 1 nm的微孔为主 ,占总孔容 ( 0 .1 65 cm3 /g)的 98% ,平均孔径 1 .83nm;Na2 CO3 制备的活性炭以 4nm的中孔为主 ,占总孔容 ( 0 .1 43cm3 /g)的 68.5 % ,平均孔径 3.42 nm,比表面积小 .3种样品的孔径都呈现出多峰分布特征 .KOH和 Na2 CO3 活化制备的活性炭的 N2 吸附脱附曲线属于 型 ,Na OH活化制备的活性炭吸附脱附曲线属于 型 .     

Electron Emission from Barium Strontium Titanate Ceramics

This paper focuses on understanding the influence of materials' properties on the ferroelectric electron emission. Ferroelectric ( x =1.0 and 0.8) and paraelectric ( x =0.67 and 0.5) compositions of barium strontium titanate (Ba _x Sr_{(1− x )}TiO₃) system were chosen for study based on their different ferroelectric and dielectric properties. Similar emission current waveforms were obtained from four compositions with negative triggering voltage applied to the rear electrode of the samples. It was difficult to explain the experimental results using the spontaneous polarization-switching model. The mechanism of electron emission from Ba _x Sr_{(1− x )}TiO₃ ceramics was ascertained to surface plasma emission.     

PVC助剂的新进展

介绍了近年来聚氯乙烯助剂的研制、开发的现状和趋势。     

Toughening by Metallic Lamina in Nickel/Alumina Composites

Tape casting and hot-pressing were used to produce metal/ alumina laminate composites. The mechanical behavior of these laminates was characterized. The strength and toughness of the laminates were greatly improved in comparison to that of monolithic alumina. Indentation strength values of the laminates remained approximately constant for indent loads between 10 and 100 N. Thus, this type of laminate appears to be a good candidate for damage-tolerant composite material design.     

NBR／EPDM的增容共混研究

本文采用第三组份增容途径,改善了NBR/EPDM并用胶的相容性、加工性和力学性能。通过对不同增容体系的研究对比,发现不同增容剂的增容效果差异很大,其巾以EVA-14增容效果最为显著,加入该增容剂的并用胶,在适当的工艺条件及硫化体系下,可获得最佳力学性能。本文还研究了不同增容剂对并用炭黑胶各种性能的影响,可以做到在不降低并用胶的力学性能的同时,提高并用胶的耐热空气老化、耐大气老化和耐臭氧老化性能.透射电镜实验和动态力学谱图证明了加与未加增容剂时,NBR/EPDM并用胶均为两相结构,增容剂起表面活性剂的作用,加强了两相界面间的相互作用,使分散相EPDM细微化和均匀化。     

PVC／PU合金的制备与研究

叙述了选择混炼型聚氨酯(PU)与分子量适中的Ⅱ型聚氯乙烯(PVC)树脂制备PVC/PU合金的过程,讨论了PVC/PU共混比、填充剂、硫化体系、共混工艺等因素对PVC/PU合金性能的影响。结果表明:选择PVC/PU=60/40(质量比),硫化剂TD15份,白炭黑30份,混炼温度140—160℃,可制得性能优良的PVC/PU合金。动态力学性能实验显示了PVC/PU合金有较好的相容性。     

Point-based online value iteration algorithm in large POMDP

Partially observable Markov decision process (POMDP) is an ideal framework for sequential decision-making under uncertainty in stochastic domains. However, it is notoriously computationally intractable to solving POMDP in real-time system. In order to address this problem, this paper proposes a point-based online value iteration (PBOVI) algorithm which involves performing value backup at specific reachable belief points, rather than over the entire belief simplex, to speed up computation processes, exploits branch-and-bound pruning approach to prune the AND/OR tree of belief states online, and proposes a novel idea to reuse the belief states that have been searched to avoid repeated computation. The experiment and simulation results show that the proposed algorithm can simultaneously satisfy the requirement of low errors and high timeliness in real-time system.