Search for: All records

A variety of colloidal structures observed in terrestrial experiments could also have been influenced by gravity effects (particle sedimentation, convection, etc.) It is often assumed that weightlessness simulated in a time-averaged sense by slowly rotating a specimen in a clinostat about an axis perpendicular to the gravity direction that is widely used in biological tests would reduce the effect of gravity on suspensions. Experiments on a non-buoyancy-matched suspension in flights in NASA Zero-gravity aircraft revealed that particle patterns formed in a clinostat and under normal gravity are actually similar. A requirement for matching densities between particles and a solvent severely limits possibilities to study the field-induced structuring in colloids in terrestrial experiments. Long-term microgravity in ISS offers unique opportunity to employ not density matched suspensions to explore a wide range of the mismatch of electric characteristics between particles and a solvent. We will report experimental data on the field driven structure formation in suspensions and present our approach to the development of ISS experiments. The aim is to understand mechanisms of structure formation and suggest novel routes for creating functional materials. *NASA NNX13AQ53G, NSF1832260. 

Generative adversarial networks (GANs) are a widely used framework for learning generative models. Wasserstein GANs (WGANs), one of the most successful variants of GANs, require solving a minmax optimization problem to global optimality, but are in practice successfully trained using stochastic gradient descent-ascent. In this paper, we show that, when the generator is a one-layer network, stochastic gradient descent-ascent converges to a global solution with polynomial time and sample complexity. 

We study the problem of inverting a deep generative model with ReLU activations. Inversion corresponds to finding a latent code vector that explains observed measurements as much as possible. In most prior works this is performed by attempting to solve a non-convex optimization problem involving the generator. In this paper we obtain several novel theoretical results for the inversion problem. We show that for the realizable case, single layer inversion can be performed exactly in polynomial time, by solving a linear program. Further, we show that for multiple layers, inversion is NP-hard and the pre-image set can be non-convex. For generative models of arbitrary depth, we show that exact recovery is possible in polynomial time with high probability, if the layers are expanding and the weights are randomly selected. Very recent work analyzed the same problem for gradient descent inversion. Their analysis requires significantly higher expansion (logarithmic in the latent dimension) while our proposed algorithm can provably reconstruct even with constant factor expansion. We also provide provable error bounds for different norms for reconstructing noisy observations. Our empirical validation demonstrates that we obtain better reconstructions when the latent dimension is large. 

The influence of the pore topology and polymer properties on mechanical characteristics of asymmetric polyethersulfone (PES) and symmetric polyvinylidene fluoride (PVDF) microfiltration membranes was investigated by conducting elongation, creep, stress relaxation, small-amplitude oscillatory and bubble point pressure tests. The main aspects of the membrane stress-strain curves were found to be similar despite significant differences in the pore topology and polymer properties. While the Kelvin-Voigt model for solid polymers described the membrane viscoelastic response below the transition to ductile yielding, the stress-strain curves of membranes and solid polymers above the yield point appeared to be drastically different. All tested membranes demonstrated weak strain hardening, low sensitivity to strain rate, significant elastic recovery, stress relaxation and reduction of the bubble point pressure with accumulation of plastic deformation. Therefore, tensile stresses exerted on a membrane under assembling and process conditions should be smaller than the yield stress to assure that they will not impair filter performance. The novelty of our approach is the use of models for perforated plates to evaluate membrane mechanical properties as ductile yielding for both proceeds via localized plastic deformation around pores. Presented results provide a reliable framework for development of membranes with properties tailored to applications.