This is the third paper of a series where we study the stellar population gradients (SP; ages, metallicities, $\alpha$-element abundance ratios and stellar initial mass functions) of early type galaxies (ETGs) at $z\le 0.08$ from the MaNGA-DR15 survey. In this work we focus on the S0 population and quantify how the SP varies across the population as well as with galactocentric distance. We do this by measuring Lick indices and comparing them to stellar population synthesis models. This requires spectra with high signal-to-noise which we achieve by stacking in bins of luminosity (L$_r$) and central velocity dispersion ($\sigma_0$). We find that: 1) There is a bimodality in the S0 population: S0s more massive than $3\times 10^{10}M_\odot$ show stronger velocity dispersion and age gradients (age and $\sigma_r$ decrease outwards) but little or no metallicity gradient, while the less massive ones present relatively flat age and velocity dispersion profiles, but a significant metallicity gradient (i.e. [M/H] decreases outwards). Above $2\times10^{11}M_\odot$ the number of S0s drops sharply. These two mass scales are also where global scaling relations of ETGs change slope. 2) S0s have steeper velocity dispersion profiles than fast rotating elliptical galaxies (E-FRs) of the same luminosity and velocity dispersion. The kinematic profiles and stellar population gradients of E-FRs are both more similar to those of slow rotating ellipticals (E-SRs) than to S0s, suggesting that E-FRs are not simply S0s viewed face-on. 3) At fixed $\sigma_0$, more luminous S0s and E-FRs are younger, more metal rich and less $\alpha$-enhanced. Evidently for these galaxies, the usual statement that 'massive galaxies are older' is not true if $\sigma_0$ is held fixed.

We present estimates of stellar population (SP) gradients from stacked spectra of slow (SR) and fast (FR) rotator elliptical galaxies from the MaNGA-DR15 survey. We find that: 1) FRs are $\sim 5$ Gyrs younger, more metal rich, less $\alpha$-enhanced and smaller than SRs of the same luminosity $L_r$ and central velocity dispersion $\sigma_0$. This explains why when one combines SRs and FRs, objects which are small for their $L_r$ and $\sigma_0$ tend to be younger. Their SP gradients are also different. 2) Ignoring the FR/SR dichotomy leads one to conclude that compact galaxies are older than their larger counterparts of the same mass, even though almost the opposite is true for FRs and SRs individually. 3) SRs with $\sigma_0\le 250$ km s$^{-1}$ are remarkably homogeneous within $\sim R_e$: they are old, $\alpha$-enhanced and only slightly super-solar in metallicity. These SRs show no gradients in age and $M_*/L_r$, negative gradients in metallicity, and slightly positive gradients in [$\alpha$/Fe] (the latter are model dependent). SRs with $\sigma_0\ge 250$ km $s^{-1}$ are slightly younger and more metal rich, contradicting previous work suggesting that age increases with $\sigma_0$. They also show larger $M_*/L_r$ gradients. 4) Self-consistently accounting for $M_*/L$ gradients yields $M_{\rm dyn}\approx M_*$ because gradients reduce $M_{\rm dyn}$ by $\sim 0.2$ dex while only slightly increasing the $M_*$ inferred using a Kroupa (not Salpeter) IMF. 5) The FR population all but disappears above $M_*\ge 3\times 10^{11}M_\odot$; this is the same scale at which the size-mass correlation and other scaling relations change. Our results support the finding that this is an important mass scale which correlates with the environment and above which mergers matter.

We estimate ages, metallicities, $\alpha$-element abundance ratios and stellar initial mass functions of elliptical (E) and S0 galaxies from the MaNGA-DR15 survey. We stack spectra and use a variety of single stellar population synthesis models to interpret the absorption line strengths in these spectra. We quantify how these properties vary across the population, as well as with galactocentric distance. This paper is the first of a series and is based on a sample of pure elliptical galaxies at z $\le$ 0.08. We show that the properties of the inner regions of Es with the largest luminosity (L$_r$) and central velocity dispersion ($\sigma_0$) are consistent with those associated with the commonly used Salpeter IMF, whereas a Kroupa-like IMF is a better description at $\sim$ 0.8R/Re (assuming [Ti/Fe] variations are limited). For these galaxies the stellar mass-to-light ratio decreases at most by a factor of 2 from the central regions to Re. In contrast, for lower L$_r$ and $\sigma_0$ galaxies, the IMF is shallower and M$_{*}$/L$_r$ in the central regions is similar to the outskirts. Although a factor of 2 is smaller than previous reports based on a handful of galaxies, it is still large enough to matter for dynamical mass estimates. Accounting self-consistently for these gradients when estimating both M$_{*}$ and M$_{dyn}$ brings the two into good agreement: gradients reduce M$_{dyn}$ by $\sim$ 0.2 dex while only slightly increasing the M$_{*}$ inferred using a Kroupa IMF. This is a different resolution of the M$_{*}$-M$_{dyn}$ discrepancy than has been followed in the recent literature where M$_{*}$ of massive galaxies is increased by adopting a Salpeter IMF while leaving Mdyn unchanged. A companion paper discusses how stellar population differences are even more pronounced if one separates slow from fast rotators.

We present a sample of 33 Balmer Break Galaxies (BBGs) selected as HST/F160W dropouts in the deepest CANDELS/GOODS fields ($H\gtrsim27.3$~mag) but relatively bright in {\it Spitzer}/IRAC ($[3.6],[4.5]<24.5$~mag), implying red colors (median and quartiles: $\langle H-[3.6]\rangle=3.1^{3.4}_{2.8}$\,mag). Half of these BBGs are newly identified sources. Our BBGs are massive ($\langle \log(\rm{M}/\rm{M}_\odot)\rangle=10.8$) high redshift ($\langle z\rangle=4.8$) dusty ($\langle \rm{A(V)}\rangle=2.0$~mag) galaxies. The SEDs of half of our sample indicate that they are star-forming galaxies with typical specific SFRs 0.5-1.0~Gyr$^{-1}$, qualifying them as main sequence (MS) galaxies at $3<z<6$. One third of those SEDs indicates the presence of prominent emission lines (H$\beta$+$[OIII]$, H$\alpha$$+$[NII]) boosting the IRAC fluxes and red colors. Approximately 20\% of the BBGs are very dusty ($\rm{A(V)}\sim2.5$~mag) starbursts with strong mid-to-far infrared detections and extreme SFRs ($\rm{SFR}>10^{3}\,\rm{M}_\odot/yr$) that place them above the MS. The rest, 30\%, are post-starbursts or quiescent galaxies located $>2\sigma$ below the MS with mass-weighted ages older than 700~Myr. Only 2 of the 33 galaxies are X-ray detected AGN with optical/near-infrared SEDs dominated by stellar emission, but the presence of obscured AGN in the rest of sources cannot be discarded. Our sample accounts for 8\% of the total number density of $\log(\rm{M}/\rm{M}_\odot)>10$ galaxies at $z>3$, but it is a significant contributor (30\%) to the general population of red $\log(\rm{M}/\rm{M}_\odot)>11$ galaxies at $4<z<6$. Finally, our results point out that 1 of every 30 massive $\log(\rm{M}/\rm{M}_\odot)>11$ galaxies in the local Universe was assembled in the first 1.5~Gyr after the Big Bang, a fraction that is not reproduced by state-of-the-art galaxy formation simulations.

Deep Learning (DL) algorithms for morphological classification of galaxies have proven very successful, mimicking (or even improving) visual classifications. However, these algorithms rely on large training samples of labelled galaxies (typically thousands of them). A key question for using DL classifications in future Big Data surveys is how much of the knowledge acquired from an existing survey can be exported to a new dataset, i.e. if the features learned by the machines are meaningful for different data. We test the performance of DL models, trained with Sloan Digital Sky Survey (SDSS) data, on Dark Energy survey (DES) using images for a sample of $\sim$5000 galaxies with a similar redshift distribution to SDSS. Applying the models directly to DES data provides a reasonable global accuracy ($\sim$ 90%), but small completeness and purity values. A fast domain adaptation step, consisting in a further training with a small DES sample of galaxies ($\sim$500-300), is enough for obtaining an accuracy > 95% and a significant improvement in the completeness and purity values. This demonstrates that, once trained with a particular dataset, machines can quickly adapt to new instrument characteristics (e.g., PSF, seeing, depth), reducing by almost one order of magnitude the necessary training sample for morphological classification. Redshift evolution effects or significant depth differences are not taken into account in this study.

We describe the SDSS-IV MaNGA PyMorph Photometric (MPP-VAC) and MaNGA Deep Learning Morphology (MDLM-VAC) Value Added Catalogs. The MPP-VAC provides photometric parameters from S\'ersic and S\'ersic+Exponential fits to the 2D surface brightness profiles of the MaNGA DR15 galaxy sample. Compared to previous PyMorph analyses of SDSS imaging, our analysis of the MaNGA DR15 incorporates three improvements: the most recent SDSS images; modified criteria for determining bulge-to-disk decompositions; and the fits in MPP-VAC have been eye-balled, and re-fit if necessary, for additional reliability. A companion catalog, the MDLM-VAC, provides Deep Learning-based morphological classifications for the same galaxies. The MDLM-VAC includes a number of morphological properties (e.g., a TType, and a finer separation between elliptical and S0 galaxies). Combining the MPP- and MDLM-VACs allows to show that the MDLM morphological classifications are more reliable than previous work. It also shows that single-S\'ersic fits to late- and early-type galaxies are likely to return S\'ersic indices of $n \le 2$ and $\ge 4$, respectively, and this correlation between $n$ and morphology extends to the bulge component as well. While the former is well-known, the latter contradicts some recent work suggesting little correlation between $n$-bulge and morphology. Combining both VACs with MaNGA's spatially resolved spectroscopy allows us to study how the stellar angular momentum depends on morphological type. We find correlations between stellar kinematics, photometric properties, and morphological type even though the spectroscopic data played no role in the construction of the MPP- and MDLM-VACs.

We have undertaken a comprehensive search for both Lyman Alpha Emitters (LAEs) and Lyman Break Galaxies (LBGs) in the SHARDS Survey of the GOODS-N field. SHARDS is a deep imaging survey, made with the 10.4 m Gran Telescopio Canarias (GTC), employing 25 medium band filters in the range from 500 to 941 nm. This is the first time that both LAEs and LBGs are surveyed simultaneously in a systematic way in a large field. We draw a sample of 1558 sources; 528 of them are LAEs. Most of the sources (1434) show rest-frame UV continua. A minority of them (124) are pure LAEs with virtually no continuum detected in SHARDS. We study these sources from $z\sim3.35$ up to $z\sim6.8$, well into the epoch of reionization. Note that surveys done with just one or two narrow band filters lack the possibility to spot the rest-frame UV continuum present in most of our LAEs. We derive redshifts, Star Formation Rates (SFRs), Ly$\alpha$ Equivalent Widths (EWs) and Luminosity Functions (LFs). Grouping within our sample is also studied, finding 92 pairs or small groups of galaxies at the same redshift separated by less than 60 comoving kpc. In addition, we relate 87 and 55 UV-selected objects with two known overdensities at $z=4.05$ and $z=5.198$, respectively. Finally, we show that surveys made with broad band filters are prone to introduce many unwanted sources ($\sim20$% interlopers), which means that previous studies may be overestimating the calculated LFs, specially at the faint end.

The physical processes driving the evolution of star formation (SF) in galaxies over cosmic time still present many open questions. Recent galaxy surveys allow now to study these processes in great detail at intermediate redshift. In this work, we build a complete sample of star-forming galaxies and analyze their properties, reaching systems with low stellar masses and low star formation rates (SFRs) at intermediate-to-low redshift. We use data from the SHARDS multiband survey in the GOODS-North field. Its depth (up to magnitude $\langle m_{3\sigma}\rangle\sim26.5$) and its spectro-photometric resolution ($R\sim50$) provides us with an ideal dataset to search for emission line galaxies (ELGs). We develop a new algorithm to identify low-redshift ($z$<0.36) ELGs by detecting the [OIII]5007 and $H\alpha$ emission lines simultaneously. We fit the spectral energy distribution (SED) of the selected sample, using a model with two single stellar populations. We find 160 star-forming galaxies with equivalent widths (EWs) as low as 12 {\AA}, with median values for the sample of $\sim$ 35 {\AA} in [OIII]5007 and $\sim$ 56 {\AA} in $H\alpha$, respectively. Results from the SED fitting show a young stellar population with low median metallicity (36% of the solar value) and extinction ($A_V \sim$ 0.37), with median galaxy stellar mass $\sim$ 10$^{8.5}$ M$_{\odot}$. Gas-phase metallicities measured from available spectra are also low. ELGs in our sample present bluer colors in the UVJ plane than the median color-selected star-forming galaxy in SHARDS. We suggest a new (V-J) color criterion to separate ELGs from non-ELGs in blue galaxy samples. In addition, several galaxies present high densities of O-type stars. Robust fits to the full SEDs can only be obtained including an old stellar population, suggesting the young component is built up by a recent burst of SF in an otherwise old galaxy.

We present a morphological catalogue for $\sim$ 670,000 galaxies in the Sloan Digital Sky Survey in two flavours: T-Type, related to the Hubble sequence, and Galaxy Zoo 2 (GZ2 hereafter) classification scheme. By combining accurate existing visual classification catalogues with machine learning, we provide the largest and most accurate morphological catalogue up to date. The classifications are obtained with Deep Learning algorithms using Convolutional Neural Networks (CNNs). We use two visual classification catalogues, GZ2 and Nair & Abraham (2010), for training CNNs with colour images in order to obtain T-Types and a series of GZ2 type questions (disk/features, edge-on galaxies, bar signature, bulge prominence, roundness and mergers). We also provide an additional probability enabling a separation between pure elliptical (E) from S0, where the T-Type model is not so efficient. For the T-Type, our results show smaller offset and scatter than previous models trained with support vector machines. For the GZ2 type questions, our models have large accuracy (> 97\%), precision and recall values (> 90\%) when applied to a test sample with the same characteristics as the one used for training. The catalogue is publicly released with the paper.

Numerous ongoing and future large area surveys (e.g. DES, EUCLID, LSST, WFIRST), will increase by several orders of magnitude the volume of data that can be exploited for galaxy morphology studies. The full potential of these surveys can only be unlocked with the development of automated, fast and reliable analysis methods. In this paper we present DeepLeGATo, a new method for two-dimensional photometric galaxy profile modeling, based on convolutional neural networks. Our code is trained and validated on analytic profiles (HST/CANDELS F160W filter) and it is able to retrieve the full set of parameters of one- component S\'ersic models: total magnitude, effective radius, S\'ersic index, axis ratio. We show detailed comparisons between our code and GALFIT. On simulated data, our method is more accurate than GALFIT and 3000 time faster on GPU (50 times when run on the same CPU). On real data, DeepLeGATo trained on simulations behaves similarly to GALFIT on isolated galaxies. With a fast domain adaptation step made with the 0.1 - 0.8 per cent the size of the training set, our code is easily capable to reproduce the results obtained with GALFIT even on crowded regions. DeepLeGATo does not require any human intervention beyond the training step, rendering it much automated than traditional profiling methods. The development of this method for more complex models (two-component galaxies, variable PSF, dense sky regions) could constitute a fundamental tool in the era of big data in astronomy.

We present Star Formation Histories (SFHs) for a sample of 104 massive (stellar mass M > 10$^{10}$ M$_{\odot}$) quiescent galaxies (MQGs) at $z$=1.0-1.5 from the analysis of spectro-photometric data from the SHARDS and HST/WFC3 G102 and G141 surveys of the GOODS-N field, jointly with broad-band observations from ultraviolet (UV) to far-infrared (Far-IR). The sample is constructed on the basis of rest-frame UVJ colours and specific star formation rates (sSFR=SFR/Mass). The Spectral Energy Distributions (SEDs) of each galaxy are compared to models assuming a delayed exponentially declining SFH. A Monte Carlo algorithm characterizes the degeneracies, which we are able to break taking advantage of the SHARDS data resolution, by measuring indices such as MgUV and D4000. The population of MQGs shows a duality in their properties. The sample is dominated (85%) by galaxies with young mass-weighted ages, tM $<$ 2 Gyr, short star formation timescales, $\langle \tau \rangle$ $\sim$ 60-200 Myr, and masses log(M/M$_{\odot}$) $\sim$ 10.5. There is an older population (15%) with tM$=$2 - 4 Gyr, longer star formation timescales, $\langle \tau \rangle$$\sim$ 400 Myr, and larger masses, log(M/M$_{\odot}$) $\sim$ 10.7. The SFHs of our MQGs are consistent with the slope and the location of the Main Sequence (MS) of star-forming galaxies at $z$ $>$ 1.0, when our galaxies were 0.5-1.0 Gyr old. According to these SFHs, all the MQGs experienced a Luminous Infrared Galaxy (LIRG) phase that lasts for $\sim$ 500 Myr, and half of them an Ultra Luminous Infrared Galaxy (ULIRG) phase for $\sim$ 100 Myr. We find that the MQG population is almost assembled at $z$ $\sim$ 1, and continues evolving passively with few additions to the population.

In this paper, we present a comprehensive analysis of star-forming galaxies (SFGs) at intermediate redshifts (z~1). We combine the ultra-deep optical spectro-photometric data from the Survey for High-z Absorption Red and Dead Sources (SHARDS) with deep UV-to-FIR observations in the GOODS-N field. Exploiting two of the 25 SHARDS medium-band filters, F687W17 and F823W17, we select [OII] emission line galaxies at z~0.84 and z~1.23 and characterize their physical properties. Their rest-frame equivalent widths (EW$_{\mathrm{rf}}$([OII])), line fluxes, luminosities, star formation rates (SFRs) and dust attenuation properties are investigated. The evolution of the EW$_{\mathrm{rf}}$([OII]) closely follows the SFR density evolution of the universe, with a trend of EW$_{\mathrm{rf}}$([OII])$\propto$(1+z)$^3$ up to redshift z~1, followed by a possible flattening. The SF properties of the galaxies selected on the basis of their [OII] emission are compared with complementary samples of SFGs selected by their MIR and FIR emission, and also with a general mass-selected sample of galaxies at the same redshifts. We demonstrate observationally that the UVJ diagram (or, similarly, a cut in the specific SFR) is only partially able to distinguish the quiescent galaxies from the SFGs. The SFR-M$_*$ relation is investigated for the different samples, yelding a logarithmic slope ~1, in good agreement with previous results. The dust attenuations derived from different SFR indicators (UV(1600), UV(2800), [OII], IR) are compared and show clear trends with respect to both the stellar mass and total SFR, with more massive and highly star-forming galaxies being affected by stronger dust attenuation.

We explore the stellar initial mass function (IMF) of a sample of 49 massive quiescent galaxies (MQGs) at 0.9$<$z$<$1.5. We base our analysis on intermediate resolution spectro-photometric data in the GOODS-N field taken in the near-infrared and optical with the HST/WFC3 G141 grism and the Survey for High-z Absorption Red and Dead Sources (SHARDS). To constrain the slope of the IMF, we have measured the TiO$_2$ spectral feature, whose strength depends strongly on the content of low-mass stars, as well as on stellar age. Using ultraviolet to near-infrared individual and stacked spectral energy distributions, we have independently estimated the stellar ages of our galaxies. Knowing the age of the stellar population, we interpret the strong differences in the TiO$_2$ feature as an IMF variation. In particular, for the heaviest z$\sim$1 MQGs (M$>$10$^{11}$Msun) we find an average age of 1.7$\pm$0.3 Gyr and a bottom-heavy IMF ($\Gamma_b$=3.2$\pm$0.2). Lighter MQGs (2$\times$10$^{10}$$<$M$<$10$^{11}$ Msun) at the same redshift are younger on average (1.0$\pm$0.2 Gyr) and present a shallower IMF slope ($\Gamma_b=2.7^{+0.3}_{-0.4}$). Our results are in good agreement with the findings about the IMF slope in early-type galaxies of similar mass in the present-day Universe. This suggests that the IMF, a key characteristic of the stellar populations in galaxies, is bottom-heavier for more massive galaxies and has remained unchanged in the last $\sim$8 Gyr.

We study a hundred of galaxies from the spectroscopic Sloan Digital Sky Survey with individual detections in the Far-Infrared Herschel PACS bands (100 or 160 $\mu$m) and in the GALEX Far-UltraViolet band up to z$\sim$0.4 in the COSMOS and Lockman Hole fields. The galaxies are divided into 4 spectral and 4 morphological types. For the star forming and unclassifiable galaxies we calculate dust extinctions from the UV slope, the H$\alpha$/H$\beta$ ratio and the $L_{\rm IR}/L_{\rm UV}$ ratio. There is a tight correlation between the dust extinction and both $L_{\rm IR}$ and metallicity. We calculate SFR$_{total}$ and compare it with other SFR estimates (H$\alpha$, UV, SDSS) finding a very good agreement between them with smaller dispersions than typical SFR uncertainties. We study the effect of mass and metallicity, finding that it is only significant at high masses for SFR$_{H\alpha}$. For the AGN and composite galaxies we find a tight correlation between SFR and L$_{IR}$ ($\sigma\sim$0.29), while the dispersion in the SFR - L$_{UV}$ relation is larger ($\sigma\sim$0.57). The galaxies follow the prescriptions of the Fundamental Plane in the M-Z-SFR space.

We analyse the rest-frame UV-to-NIR spectral energy distribution (SED) of Lyman break galaxies (LBGs), star-forming (SF) BzK (sBzK), and UV-selected galaxies at 1.5 < z < 2.5 in the COSMOS, GOODS-N, and GOODS-S fields. Additionally, we complement the multi-wavelength coverage of the galaxies located in the GOODS fields with deep FIR data taken from the GOODS-Herschel project. We find that sBzK galaxies represent the general population of SF galaxies at z ~ 2 better than LBGs. For a given stellar mass, LBGs tend to have bluer optical colours than sBzK and UV-selected galaxies. We find clean PACS individual detections for a subsample of 48 LBGs, 89 sBzK, and 91 UV-selected galaxies, that measure their dust emission directly. Their SFR_total = SFR_UV + SFR_IR cannot be recovered with the dust-correction factors derived with their continuum slope and the IRX-beta relations for local starbursts. This has implications, for example, in the definition of the main sequence (MS) at z ~ 2. PACS-detected galaxies are located above the z ~ 2 MS and thus their star formation is probably driven by starburst. This is in agreement with the shape of their IR SEDs. PACS-detected galaxies with redder UV continuum slope and higher stellar mass are more attenuated. We find that for a given UV continuum slope the dustiest galaxies at higher redshifts are more attenuated and that for a given stellar mass the dustiest galaxies at higher redshifts have stronger FIR emission. This suggests an evolution of their dust properties. However, we do not find significant evolution in the relation between dust attenuation and stellar mass with redshift, at least at z < 2.5. There is a subpopulation of 17, 26, and 27 LBGs, sBzK, and UV-selected galaxies, respectively, that are detected in any of the SPIRE bands. We speculate that SPIRE-detected LBGs are the bridging population between sub-mm galaxies and LBGs.

Lyman break galaxies (LBGs) represent one of the kinds of star-forming galaxies that are found in the high-redshift universe. The detection of LBGs in the FIR domain can provide very important clues on their dust attenuation and total SFR, allowing a more detailed study than those performed so far. In this work we explore the FIR emission of a sample of 16 LBGs at z ~ 3 in the GOODS-North and GOODS-South fields that are individually detected in PACS-100um or PACS-160um. These detections demonstrate the possibility of measuring the dust emission of LBGs at high redshift. We find that PACS-detected LBGs at z ~ 3 are highly obscured galaxies which belong to the Ultra luminous IR galaxies or Hyper luminous IR galaxies class. Their total SFR cannot be recovered with the dust attenuation factors obtained from their UV continuum slope or their SED-derived dust attenuation employing Bruzual & Charlot (2003) templates. Both methods underestimate the results for most of the galaxies. Comparing with a sample of PACS-detected LBGs at z ~ 1 we find evidences that the FIR emission of LBGs might have changed with redshift in the sense that the dustiest LBGs found at z ~ 3 have more prominent FIR emission, are dustier for a given UV slope, and have higher SFR for a given stellar mass than the dustiest LBGs found at z ~ 1.

We have studied the evolution of high redshift quiescent galaxies over an effective area of ~1.7 deg^2 in the COSMOS field. Galaxies have been divided according to their star-formation activity and the evolution of the different populations has been investigated in detail. We have studied an IRAC (mag_3.6 < 22.0) selected sample of ~18000 galaxies at z > 1.4 with multi-wavelength coverage. We have derived accurate photometric redshifts (sigma=0.06) and other important physical parameters through a SED-fitting procedure. We have divided our sample into actively star-forming, intermediate and quiescent galaxies depending on their specific star formation rate. We have computed the galaxy stellar mass function of the total sample and the different populations at z=1.4-3.0. We have studied the properties of high redshift quiescent galaxies finding that they are old (1-4 Gyr), massive (log(M/M_sun)~10.65), weakly star forming stellar populations with low dust extinction (E(B-V) < 0.15) and small e-folding time scales (tau ~ 0.1-0.3 Gyr). We observe a significant evolution of the quiescent stellar mass function from 2.5 < z < 3.0 to 1.4 < z < 1.6, increasing by ~ 1 dex in this redshift interval. We find that z ~ 1.5 is an epoch of transition of the GSMF. The fraction of star-forming galaxies decreases from 60% to 20% from z ~ 2.5-3.0 to z ~ 1.4-1.6 for log(M/M_sun) > 11, while the quiescent population increases from 10% to 50% at the same redshift and mass intervals. We compare the fraction of quiescent galaxies derived with that predicted by theoretical models and find that the Kitzbichler & White (2007) model is the one that better reproduces the data. Finally, we calculate the stellar mass density of the star-forming and quiescent populations finding that there is already a significant number of quiescent galaxies at z > 2.5 (rho~6.0 MsunMpc^-3).