Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userUnsupervised feature-learning for galaxy SEDs with denoising autoencoders
133
0
0.0
(
0
)
Added by
Joana Frontera Pons
Publication date
2017
fields
Physics
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
With the increasing number of deep multi-wavelength galaxy surveys, the spectral energy distribution (SED) of galaxies has become an invaluable tool for studying the formation of their structures and their evolution. In this context, standard analysis relies on simple spectro-photometric selection criteria based on a few SED colors. If this fully supervised classification already yielded clear achievements, it is not optimal to extract relevant information from the data. In this article, we propose to employ very recent advances in machine learning, and more precisely in feature learning, to derive a data-driven diagram. We show that the proposed approach based on denoising autoencoders recovers the bi-modality in the galaxy population in an unsupervised manner, without using any prior knowledge on galaxy SED classification. This technique has been compared to principal component analysis (PCA) and to standard color/color representations. In addition, preliminary results illustrate that this enables the capturing of extra physically meaningful information, such as redshift dependence, galaxy mass evolution and variation over the specific star formation rate. PCA also results in an unsupervised representation with physical properties, such as mass and sSFR, although this representation separates out. less other characteristics (bimodality, redshift evolution) than denoising autoencoders.
rate research
Read More
We use spatially-sparse two, three and four dimensional convolutional autoencoder networks to model sparse structures in 2D space, 3D space, and 3+1=4 dimensional space-time. We evaluate the resulting latent spaces by testing their usefulness for downstream tasks. Applications are to handwriting recognition in 2D, segmentation for parts in 3D objects, segmentation for objects in 3D scenes, and body-part segmentation for 4D wire-frame models generated from motion capture data.
High resolution galaxy spectra contain much information about galactic physics, but the high dimensionality of these spectra makes it difficult to fully utilize the information they contain. We apply variational autoencoders (VAEs), a non-linear dimensionality reduction technique, to a sample of spectra from the Sloan Digital Sky Survey. In contrast to Principal Component Analysis (PCA), a widely used technique, VAEs can capture non-linear relationships between latent parameters and the data. We find that a VAE can reconstruct the SDSS spectra well with only six latent parameters, outperforming PCA with the same number of components. Different galaxy classes are naturally separated in this latent space, without class labels having been given to the VAE. The VAE latent space is interpretable because the VAE can be used to make synthetic spectra at any point in latent space. For example, making synthetic spectra along tracks in latent space yields sequences of realistic spectra that interpolate between two different types of galaxies. Using the latent space to find outliers may yield interesting spectra: in our small sample, we immediately find unusual data artifacts and stars misclassified as galaxies. In this exploratory work, we show that VAEs create compact, interpretable latent spaces that capture non-linear features of the data. While a VAE takes substantial time to train (~1 day for 48000 spectra), once trained, VAEs can enable the fast exploration of large astronomical data sets.
We present a new method to classify the broad band optical-NIR spectral energy distributions (SEDs) of galaxies using three shape parameters (super-colours) based on a Principal Component Analysis of model SEDs. As well as providing a compact representation of the wide variety of SED shapes, the method allows for easy visualisation of information loss and biases caused by the incomplete sampling of the rest-frame SED as a function of redshift. We apply the method to galaxies in the UKIDSS Ultra Deep Survey with 0.9<z<1.2, and confirm our classifications by stacking rest-frame optical spectra for a fraction of objects in each class. As well as cleanly separating a tight red-sequence from star-forming galaxies, three unusual populations are identifiable by their unique colours: very dusty star-forming galaxies with high metallicity and old mean stellar age; post-starburst galaxies which have formed greater than around 10% of their mass in a recent unsustained starburst event; and metal-poor quiescent dwarf galaxies. We find that quiescent galaxies account for 45% of galaxies with log(M*/Msol)>11, declining steadily to 13% at log(M*/Msol)=10. The properties and mass-function of the post-starburst galaxies are consistent with a scenario in which gas-rich mergers contribute to the growth of the low and intermediate mass range of the red sequence.
Establishing accurate morphological measurements of galaxies in a reasonable amount of time for future big-data surveys such as EUCLID, the Large Synoptic Survey Telescope or the Wide Field Infrared Survey Telescope is a challenge. Because of its high level of abstraction with little human intervention, deep learning appears to be a promising approach. Deep learning is a rapidly growing discipline that models high-level patterns in data as complex multilayered networks. In this work we test the ability of deep convolutional networks to provide parametric properties of Hubble Space Telescope like galaxies (half-light radii, Sersic indices, total flux etc..). We simulate a set of galaxies including point spread function and realistic noise from the CANDELS survey and try to recover the main galaxy parameters using deep-learning. We com- pare the results with the ones obtained with the commonly used profile fitting based software GALFIT. This way showing that with our method we obtain results at least equally good as the ones obtained with GALFIT but, once trained, with a factor 5 hundred time faster.
In this paper, we propose an instance similarity learning (ISL) method for unsupervised feature representation. Conventional methods assign close instance pairs in the feature space with high similarity, which usually leads to wrong pairwise relationship for large neighborhoods because the Euclidean distance fails to depict the true semantic similarity on the feature manifold. On the contrary, our method mines the feature manifold in an unsupervised manner, through which the semantic similarity among instances is learned in order to obtain discriminative representations. Specifically, we employ the Generative Adversarial Networks (GAN) to mine the underlying feature manifold, where the generated features are applied as the proxies to progressively explore the feature manifold so that the semantic similarity among instances is acquired as reliable pseudo supervision. Extensive experiments on image classification demonstrate the superiority of our method compared with the state-of-the-art methods. The code is available at https://github.com/ZiweiWangTHU/ISL.git.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
