We consider small-scale spheroidal clusters of weakly interacting massive particles in our Galaxy as non-compact gravitational microlenses and predict the appearance of caustics in the plane of a lensed source. The crossing of these caustics by a lensed star can produce a large variety of light curves, including some observed in actual microlensing events that have been interpreted as manifestations of binary gravitational lenses. We consider also observable effects during the gravitational microlensing of stars of non-zero angular size with a given brightness distribution across their disks by such an exotic objects as natural wormholes and objects whose space-time environment is described with the NUT metric. We demonstrate that, under certain conditions, the microlensing light curves, chromatic and polarizational effects due to the properties of the lens and the star disk brightness distributions can differ considerably from those observed for a Schwarzschild gravitational lens, so that their analysis can facilitate the identification of such objects.

We carried out numerical experiments on the evaluation of the possibilities of obtaining the information about brightness distributions for the components of eclipsing variables from the data of high-precision photometry expected for planned satellites COROT and Kepler. We examined a simple model of the eclipsing binary with the spherical components on circular orbits and the linear law of the limb darkening. The solutions of light curves have been obtained as by fitting of the nonlinear model, into the number of parameters of which included the limb darkening coefficients, so also by the solution of the ill-posed inverse problem of restoration of brightness distributions across the disks of stars without rigid model constraints on the form of these functions. The obtained estimations show that if the observational accuracy amounts to 0.0001 then the limb darkening coefficients can be found with the relative error approximately 0.01 . The brightness distributions across the disks of components can be restored also nearly with the same accuracy.

We present a technique for the successive restoration of the branches of the one-dimensional strip brightness distribution across a quasar's accretion disk via the analysis of observations of high magnification events in measured fluxes from the multiple quasar images produced by a gravitational lens. Hypothesizing these events to by associated with microlensing by a fold caustic, the branches of brightness distribution are searched for on compact sets of non-negative, monotonically non-increasing, convex downward functions. The results of numerical simulations show that the solution obtained is stable against random noise. Analysis of the light curves of high magnification events in the fluxes from components C and A of the gravitational lens QSO 2237+0305, observed by the OGLE and GLITP groups, has yielded the forms of the strip brightness distributions across the accretion disk of the lensed quasar. The resulting sizes of the accretion disk are in agreement with results obtained earlier via model-fitting. The form of the brightness distribution is consistent with the expected appearance of an accretion disk rotating around supermassive black hole.

For an analysis of microlensing observational data in case of binary gravitational lenses as well as for an interpretation of observations of high magnification events in multiple images of a lensed quasar it is necessary to calculate for a given source the microlensing light curve by a fold caustic. This problem comes to the numerical calculation of a singular integral. We formulated the sufficient condition of a convergence of the integral sum for this singular integral. The strictly approach to the problem of a comparison of model results with the unequally sampled observational data consists in calculation of the model light curve in equidistant points of the canonical dissection of the integration segment and a following interpolation of its values at the moments of observations.

The problem of restoration of the source brightness distribution from an analysis of the stellar and AGNs microlensing light curves is investigated. In case of microlensing of stars by a point-mass lens as well as for caustic crossing events for binary lens the problem can be reduced to solution of the Fredholm integral equation of the 1st kind. Concrete form of the kernel of this equation depends on a type of the microlensing event. Assuming the circular symmetry of the stellar disk the search for radial brightness distribution can be carried out in the special compact sets of functions which correspond to the physics of the problem. These sets include the non-negative functions that are not increasing with increasing distance from the center of stellar disk and the upwards convex non-negative functions. The brightness distribution for the AGNs accretion disks is also circularly symmetric, but only in the locally co-moving frame. Therefore, the kernel of integral equation that determined the AGN microlensing light curve must take into account equally with the projection effect on picture plane the influence of relativistic effects. The search for solution of this equation can be carried out in the set of non-negative down convex functions. The results of analysis of microlensing light curves for the red giant MACHO Alert 95-30 and the A6 star MACHO 98-SMC-1 as well as the results of numerical simulations for the AGN microlensing observations are given.