Context. Convolutional neural networks (CNNs) have been proven to perform fast classification and detection on natural images and have potential to infer astrophysical parameters on the exponentially increasing amount of sky survey imaging data. The inference pipeline can be trained either from real human-annotated data or simulated mock observations. Until now star cluster analysis was based on integral or individual resolved stellar photometry. This limits the amount of information that can be extracted from cluster images. Aims. Develop a CNN-based algorithm aimed to simultaneously derive ages, masses, and sizes of star clusters directly from multi-band images. Demonstrate CNN capabilities on low mass semi-resolved star clusters in a low signal-to-noise ratio regime. Methods. A CNN was constructed based on the deep residual network (ResNet) architecture and trained on simulated images of star clusters with various ages, masses, and sizes. To provide realistic backgrounds, M31 star fields taken from the PHAT survey were added to the mock cluster images. Results. The proposed CNN was verified on mock images of artificial clusters and has demonstrated high precision and no significant bias for clusters of ages $\lesssim$3Gyr and masses between 250 and 4,000 ${\rm M_\odot}$. The pipeline is end-to-end, starting from input images all the way to the inferred parameters; no hand-coded steps have to be performed: estimates of parameters are provided by the neural network in one inferential step from raw images.

We present a new method to derive 2D star formation histories in dwarf irregular galaxies. Based on multicolor stellar photometry data we have found that in the Leo A galaxy during the last $\sim$400 Myr star formation was propagating according to the inside-out scenario. Star-forming regions have spread strongly asymmetrically from the center and their present day distribution correlates well with the HI surface density maps.

We have studied young stellar populations and star clusters in the dwarf irregular galaxy Leo A using multicolor ($B$, $V$, $R$, $I$, $H\alpha$) photometry data obtained with the Subaru Suprime-Cam and two-color photometry results measured on archival HST/ACS $F475W$ and $F814W$ frames. The analysis of the main sequence (MS) and blue supergiant (BSG - "blue loop") stars enabled us to study the star formation history in the Leo A galaxy during the last $\sim$200 Myr. Also, we have discovered 5 low-mass ($\lesssim$400 M$_\odot$) star clusters within the ACS field. This finding, taking into account a low metallicity environment and a yet-undetected molecular gas in Leo A, constrains star formation efficiency estimates and scenarios. Inside the well known "hole" in the HI column density map (Hunter et al. 2012) we found a shock front (prominent in $H\alpha$), implying an unseen progenitor and reminding the "hole" problems widely discussed by Warren et al. (2011).

We analysed a population of bright-red (BR) stars in the dwarf irregular galaxy Leo A by using multicolour photometry data obtained with the Subaru/Suprime-Cam ($B$, $V$, $R$, $I$, $H\alpha$) and HST/ACS ($F475W$ & $F814W$) instruments. In order to separate the Milky Way (MW) and Leo A populations of red stars, we developed a photometric method, which enabled us to study the spatial distribution of BR stars within the Leo A galaxy. We found a significant difference in the scale-length (S-L) of radial distributions of the "young" and "old" red giant branch (RGB) stars -- $0.'82 \pm 0.'04$ and $1.'53 \pm 0.'03$, respectively. Also, we determined the S-L of BR stars of $0.'85 \pm 0.'05$, which closely matches that of the "young" RGB stars. Additionally, we found a sequence of peculiar RGB stars and 8 dust-enshrouded stars in the Leo A galaxy.

Context. An automatic tool to derive structural parameters of semi-resolved star clusters located in crowded stellar fields in nearby galaxies is needed for homogeneous processing of archival frames. Aims. We have developed a program that automatically derives the structural parameters of star clusters and estimates errors by accounting for individual stars and variable sky background. Methods. Models of observed frames consist of the cluster's surface brightness distribution, convolved with a point spread function; the stars, represented by the same point spread function; and a smoothly variable sky background. The cluster's model is fitted within a large radius by using the Levenberg-Marquardt and Markov chain Monte Carlo algorithms to derive structural parameters, the flux of the cluster, and individual fluxes of all well-resolved stars. Results. FitClust, a program to derive structural parameters of semi-resolved clusters in crowded stellar fields, was developed and is available for free use. The program was tested on simulated cluster frames, and was used to measure clusters of the M31 galaxy in Subaru Suprime-Cam frames. Conclusions. Accounting for bright resolved stars and variable sky background significantly improves the accuracy of derived structural parameters of star clusters. However, their uncertainty remains dominated by the stochastic noise of unresolved stars.

We improved the stochastic 2-D galaxy disk models (Mineikis & Vansevi\v{c}ius 2014a) by introducing enriched gas outflows from galaxies and synthetic color-magnitude diagrams of stellar populations. To test the models, we use the HST/ACS stellar photometry data in four fields located along the major axis of the galaxy M33 (Williams et al. 2009) and demonstrate the potential of the models to derive 2-D star formation histories in the resolved disk galaxies.

Context. When trying to derive the star cluster physical parameters of the M33 galaxy using broad-band unresolved ground-based photometry, previous studies mainly made use of simple stellar population models, shown in the recent years to be oversimplified. Aims. In this study, we aim to derive the star cluster physical parameters (age, mass, and extinction; metallicity is assumed to be LMC-like for clusters with age below 1\,Gyr and left free for older clusters) of this galaxy using models that take stochastic dispersion of cluster integrated colors into account. Methods. We use three recently published M33 catalogs of cluster optical broad-band photometry in standard $UBVRI$ and in CFHT/MegaCam $u^{*}g'r'i'z'$ photometric systems. We also use near-infrared $JHK$ photometry that we derive from deep 2MASS images. We derive the cluster parameters using a method that takes into account the stochasticity problem, presented in previous papers of this series. Results. The derived differential age distribution of the M33 cluster population is composed of a two-slope profile indicating that the number of clusters decreases when age gets older. The first slope is interpreted as the evolutionary fading phase of the cluster magnitudes, and the second slope as the cluster disruption. The threshold between these two phases occurs at $\sim$300\,Myrs, comparable to what is observed in the M31 galaxy. We also model by use of artificial clusters the ability of the cluster physical parameter derivation method to correctly derive the two-slope profile for different photometric systems tested.

This study is the third of a series that investigates the degeneracy and stochasticity problems present in the determination of physical parameters such as age, mass, extinction, and metallicity of partially resolved or unresolved star cluster populations situated in external galaxies when using broad-band photometry. This work tests the derivation of parameters of artificial star clusters using models with fixed and free metallicity for the WFC3+ACS photometric system. Then the method is applied to derive parameters of a sample of 203 star clusters in the Andromeda galaxy observed with the HST. Following Papers I \& II, the star cluster parameters are derived using a large grid of stochastic models that are compared to the observed cluster broad-band integrated WFC3+ACS magnitudes. We derive the age, mass, and extinction of the sample of M31 star clusters with one fixed metallicity in agreement with previous studies. Using artificial tests we demonstrate the ability of the WFC3+ACS photometric system to derive the metallicity of star clusters. We show that the metallicity derived using photometry of 36 massive M31 star clusters is in a good agreement with the metallicity previously derived using spectroscopy taken from literature.

We have developed a fast numerical 2-D model of galaxy disk evolution (resolved along the galaxy radius and azimuth) by adopting a scheme of parameterized stochastic self-propagating star formation. We explore the parameter space of the model and demonstrate its capability to reproduce 1-D radial profiles of the galaxy M33: gas surface density, surface brightness in the i and GALEX FUV passbands, and metallicity.

We investigate the effects of stochasticity on the observed galaxy parameters by comparing our stochastic star formation two-dimensional (2-D) galaxy evolution models with the commonly used one-dimensional (1-D) models with smooth star formation. The 2-D stochastic models predict high variability of the star formation rate and the surface photometric parameters across the galactic disks and in time.

We have surveyed a complete extent of Leo A - an apparently isolated gas-rich low-mass dwarf irregular galaxy in the Local Group. The $B$, $V$, and $I$ passband CCD images (typical seeing $\sim$0.8") were obtained with Subaru Telescope equipped with Suprime-Cam mosaic camera. The wide-field ($20' \times 24'$) photometry catalog of 38,856 objects ($V \sim 16-26$ mag) is presented. This survey is also intended to serve as "a finding chart" for future imaging and spectroscopic observation programs of Leo A.

Stochasticity of bright stars introduces uncertainty and bias into derived structural parameters of star clusters. We have simulated a grid of cluster $V$-band images, observed with Subaru Suprime-Cam with age, mass, and size representing a cluster population in the M31 galaxy and derived their structural parameters by fitting King model to the surface brightness distribution. We have found that clusters less massive than $10^4 M_\odot$ show significant uncertainty in their core and tidal radii for all ages, while clusters younger than 10 Myr have their sizes systematically underestimated for all masses. This emphasizes the importance of stochastic simulations to asses the true uncertainty of structural parameters in studies of semi-resolved and unresolved clusters.

This paper is the second of a series that investigates the stochasticity and degeneracy problems that hinder the derivation of the age, mass, extinction, and metallicity of unresolved star clusters in external galaxies when broad-band photometry is used. While Paper I concentrated on deriving age, mass, and extinction of star clusters for one fixed metallicity, we here derive these parameters in case when metallicity is let free to vary. The results were obtained using several different filter systems ($UBVRI$, $UBVRIJHK$, GALEX+$UBVRI$), which allowed to optimally reduce the different degeneracies between the cluster physical parameters. The age, mass, and extinction of a sample of artificial star clusters were derived by comparing their broad-band integrated magnitudes with the magnitudes of a large grid of cluster models with various metallicities. A large collection of artificial clusters was studied to model the different degeneracies in the age, mass, extinction, and metallicity parameter space when stochasticity is taken into account in the cluster models. We show that, without prior knowledge on the metallicity, the optical bands ($UBVRI$) fail to allow a correct derivation of the age, mass, and extinction because of the strong degeneracies between models of different metallicities. Adding near-infrared information ($UBVRI$+$JHK$) slightly helps in improving the parameter derivation, except for the metallicity. Adding ultraviolet data (GALEX+$UBVRI$) helps significantly in deriving these parameters and allows constraining the metallicity when the photometric errors have a Gaussian distribution with standard deviations 0.05 mag for $UBVRI$ and 0.15 mag for the GALEX bands.

We developed a method for a fast modeling of broad-band UBVRI integrated magnitudes of unresolved star clusters and used it to derive their physical parameters (age, mass, and extinction). The method was applied on M33 galaxy cluster sample and consistency of ages and masses derived from unresolved observations with the values derived from resolved stellar photometry was demonstrated. We found that interstellar extinction causes minor age-extinction degeneracy for the studied sample due to a narrow extinction range in M33.

Context. Stochasticity and physical parameter degeneracy problems complicate the derivation of the parameters (age, mass, and extinction) of unresolved star clusters when using broad-band photometry. Aims. We develop a method to simulate stochasticity and degeneracies, and to investigate their influence on the accuracy of derived physical parameters. Then we apply it to star cluster samples of M31 and M33 galaxies. Methods. Age, mass and extinction of observed star clusters are derived by comparing their broad-band UBVRI integrated magnitudes to the magnitudes of a large grid of star cluster models with fixed metallicity Z=0.008. Masses of stars for a cluster model are randomly sampled from the initial mass function. Models of star clusters from the model grid, which have all of their magnitudes located within 3 observational errors from the magnitudes of the observed cluster, are selected for the computation of its age, mass, and extinction. Results. In the case of the M31 galaxy, the extinction range is wide and the age-extinction degeneracy is strong for a fraction of its clusters. Because of a narrower extinction range, the age-extinction degeneracy is weaker for the M33 clusters. By using artificial cluster sample, we show that age-extinction degeneracy can be reduced significantly if the range of intrinsic extinction within the host galaxy is narrow.

We present a model of chemical and spectrophotometric evolution of disk galaxies based on a stochastic self-propagating star formation scenario. The model incorporates galaxy formation through the process of accretion, chemical and photometric evolution treatment, based on simple stellar populations (SSP), and parameterized gas dynamics inside the model. The model reproduces observational data of a late-type spiral galaxy M33 reasonably well. Promising test results prove the applicability of the model and the adequate accuracy for the interpretation of disk galaxy properties.