We have recently witnessed tremendous success of Machine Learning (ML) in practical applications. Computer vision, speech recognition and language translation have all seen a near human level performance. We expect, in the near future, most business applications will have some form of ML. However, testing such applications is extremely challenging and would be very expensive if we follow today's methodologies. In this work, we present an articulation of the challenges in testing ML based applications. We then present our solution approach, based on the concept of Metamorphic Testing, which aims to identify implementation bugs in ML based image classifiers. We have developed metamorphic relations for an application based on Support Vector Machine and a Deep Learning based application. Empirical validation showed that our approach was able to catch 71% of the implementation bugs in the ML applications.

No abstract

We consider the slow motion of a gas bubble in a cylindrical column filled with a viscoplastic fluid, modeled here as a Herschel–Bulkley fluid. Because of the yield stress of the fluid, it is possible that a bubble will remain trapped in the fluid indefinitely. We adapt Prager’s two variational principles to our problem. From these variational principles we develop two general stopping conditions, i.e., for a given bubble we can calculate a critical Bingham number above which the bubble will not move. The first condition is derived by bounding the velocity field and the second condition by bounding the stress field. We illustrate these conditions by considering specific bubble shapes, e.g., axisymmetric bubbles. We also develop a condition for bubble motion.

Surfactant molecules can self-assemble into various morphologies under proper combinations of ionic strength, temperature, and flow conditions. At equilibrium, wormlike micelles can transition from entangled to branched and multiconnected structures with increasing salt concentration. Under certain flow conditions, micellar structural transitions follow different trajectories. In this work, we consider the flow of two semidilute wormlike micellar solutions through microposts, focusing on their microstructural and rheological evolutions. Both solutions contain cetyltrimethylammonium bromide and sodium salicylate. One is weakly viscoelastic and shear thickening, whereas the other is strongly viscoelastic and shear thinning. When subjected to strain rates of ∼10 3 s −1 and strains of ∼10 3 , we observe the formation of a stable flow-induced structured phase (FISP), with entangled, branched, and multiconnected micellar bundles, as evidenced by electron microscopy. The high stretching and flow alignment in the microposts enhance the flexibility and lower the bending modulus of the wormlike micelles. As flexible micelles flow through the microposts, it becomes energetically favorable to minimize the number of end caps while concurrently promoting the formation of cross-links. The presence of spatial confinement and extensional flow also enhances entropic fluctuations, lowering the energy barrier between states, thus increasing transition frequencies between states and enabling FISP formation. Whereas the rheological properties (zero-shear viscosity, plateau modulus, and stress relaxation time) of the shear-thickening precursor are smaller than those of the FISP, those of the shear-thinning precursor are several times larger than those of the FISP. This rheological property variation stems from differences in the structural evolution from the precursor to the FISP. microfluidics | microrheology | mesh size S urfactant molecules in aqueous solutions can self-assemble into different structures, such as spherical micelles, cylindrical micelles, lamellar phases, and vesicles (1). The morphology of these self-assembled structures is influenced by surfactant concentration, temperature, external additives (e.g., cosurfactants or salts), and flow conditions. Cylindrical micelles in the presence of inorganic or organic salts can self-assemble into large, flexible, and elongated wormlike micelles that exhibit viscoelastic properties (2-4). The ionic strength of the salt screens the electrostatic repulsions between the charged surfactant head groups, promoting cylindrical micellar growth (1-3). At higher salt concentrations, the entangled linear micelles can transition to branched and multiconnected micellar networks, as evidenced by rheological measurements, light scattering techniques, and electron microcopy (EM) imaging (5-18). Additionally, wormlike micellar solutions are known to exhibit a variety of interesting phenomena, some of which are shear banding (19-21), shear thickening (22, 23), shear-induced transitions and insta...

No abstract