Privacy-Preserving Image Template Sharing Using Contrastive Learning

With the recent developments of Machine Learning as a Service (MLaaS), various privacy concerns have been raised. Having access to the user’s data, an adversary can design attacks with different objectives, namely, reconstruction or attribute inference attacks. In this paper, we propose two different training frameworks for an image classification task while preserving user data privacy against the two aforementioned attacks. In both frameworks, an encoder is trained with contrastive loss, providing a superior utility-privacy trade-off. In the reconstruction attack scenario, a supervised contrastive loss was employed to provide maximal discrimination for the targeted classification task. The encoded features are further perturbed using the obfuscator module to remove all redundant information. Moreover, the obfuscator module is jointly trained with a classifier to minimize the correlation between private feature representation and original data while retaining the model utility for the classification. For the attribute inference attack, we aim to provide a representation of data that is independent of the sensitive attribute. Therefore, the encoder is trained with supervised and private contrastive loss. Furthermore, an obfuscator module is trained in an adversarial manner to preserve the privacy of sensitive attributes while maintaining the classification performance on the target attribute. The reported results on the CelebA dataset validate the effectiveness of the proposed frameworks.     

Active transient elasto-acoustic response damping of a thick-walled liquid-coupled piezolaminated cylindrical vessel

The linear 3D piezoelasticity theory in conjunction with the versatile transfer matrix approach and the wave equation for the internal acoustic domain are employed for active non-stationary vibroacoustic response control of an arbitrarily thick, tri-laminate, fluid-filled, simply supported, piezocomposite cylindrical tank, excited by arbitrary (non-axisymmetric) time-dependent on-surface mechanical loads. The smart structure is composed of a supporting core layer of functionally graded orthotropic material perfectly bonded to inner and outer spatially distributed radially polarized functionally graded piezoceramic sensor and uniform force actuator (FGPM) layers. Active vibration damping is implemented by transferring the accumulated voltage on the sensor layer to the piezoelectric actuator layer in context of proportional and derivative control laws. Durbin's numerical inverse Laplace transform scheme is utilized to calculate the time response histories of the relevant interface displacement/stress components, center-point acoustic pressure, and actuator voltage, for selected loading configurations (i.e., concentrated step, impulse, and moving external loads). Numerical simulations demonstrate the effectiveness of the adopted distributed sensing/actuation configuration together with the active damping control strategy in suppressing the vibroacoustic response of a three-layered (Ba₂NaNb₅O₁₅/Al/PZT4) water-filled piezoelastic cylindrical tank. Limiting cases are considered and the validity of results is established by comparison with the available data as well as with the aid of a commercial finite element package.     

Facile,one-pot synthesis of new phenanthridine derivatives through 1,4-dipolar cycloaddition of phenantridine,activated acetylenes,and aromatic aldehydes

A facile, one-pot synthesis of new phenanthridine derivatives by a three-component reaction between phenantridine, dialkyl acetylenedicarboxylates（DAADs）, and aromatic aldehydes is described. The reaction is characterized by mild conditions, high selectivity, and tolerance to various functional groups.     

Supplier quality improvement: The value of information under uncertainty

We consider supplier development decisions for prime manufacturers with extensive supply bases producing complex, highly engineered products. We propose a novel modelling approach to support supply chain managers decide the optimal level of investment to improve quality performance under uncertainty. We develop a Poisson–Gamma model within a Bayesian framework, representing both the epistemic and aleatory uncertainties in non-conformance rates. Estimates are obtained to value a supplier quality improvement activity and assess if it is worth gaining more information to reduce epistemic uncertainty. The theoretical properties of our model provide new insights about the relationship between the degree of epistemic uncertainty, the effectiveness of development programmes, and the levels of investment. We find that the optimal level of investment does not have a monotonic relationship with the rate of effectiveness. If investment is deferred until epistemic uncertainty is removed then the expected optimal investment monotonically decreases as prior variance increases but only if the prior mean is above a critical threshold. We develop methods to facilitate practical application of the model to industrial decisions by a) enabling use of the model with typical data available to major companies and b) developing computationally efficient approximations that can be implemented easily. Application to a real industry context illustrates the use of the model to support practical planning decisions to learn more about supplier quality and to invest in improving supplier capability.     

In situ crystallization of ZnAl2O4/ZnO nanocomposite on alumina granule for photocatalytic purification of wastewater

Research on Chemical Intermediates - ZnAl2O4/ZnO nanocomposites with different ZnO (20, 30, and 40 mol%) concentrations and coated samples on supports were successfully prepared through...