Mar
2018

Classifying the Large Scale Structure of the Universe with Deep Neural Networks.

What do cosmological filaments and walls have in common with soft tissue and bone? it is more than 10 years ago that I developed the Multiscale Morphology Filter (MMF) based on medical imaging techniques for vessel and bone segmentation using Hessian filters. This papers presents the first application of Deep Convolutional Neural Networks to the segmentation of cosmological filaments and walls. The main idea (U-Net) was developed for medical imaging problems.

Read the paper
Dec
2017

Predicting cosmic environment with machine learning.

We are trying to predict the cosmic environment of a galaxy based on the properties of the galaxy. This is essencially the inverse of what is traditionally done in galaxy-environment studies where galaxy properties are predicted from environment. What we learned is that the assembly history of a galaxy is sensitive to its environment so simple criteria to select galaxies like the Milky Way by mass and assembly history make sense after all.

Read the paper
Feb
2015

NebulOS, Big Data for science.

NebulOS is a Big Data platform for general computing that allows a datacenter to be treated as a single computer. With NebulOS, the process of writing a massively parallel program for a datacenter is as simple as writing a Python script for a desktop computer. Users can run pre-existing analysis software on data distributed over thousands of machines with just a few lines of python. The platform is built upon industry-standard, open-source Big Data technologies, from which it inherits fast data throughput and fault tolerance. NebulOS' main developer was Nathaniel Stickey working under my supervision as my postdoc. This work was funded by a UCR research and development grant.

Read a detailed description at Nathaniel's blog
Jun
2017

Sub-megaparsec Redshift Estimates from Cosmic Web Constraints

This is one of those ideas that seem to good to be true. Conventional photo-z techniques predict distances to galaxies with such large uncertainties that any structure in the galaxy distribution is erased (left panel above). We realized that we can use the properties of the cosmic web to impose strong constraints on the galaxies position and predict accurate distances with errors two orders of magnitude smaller than previous methods.

Read the paper
Feb
2013

The Multum in Parvo (MIP) simulation.

The MIP simulation is a special kind of ensemble simulation in which all realizations share the same large-scale structure features. A simple way of describing this is as a "parallel universe simulation". The key of the MIP is generating initial conditions that are correlated above a given scale, this ensures that each realization contains the same cosmic web structures defined by a unique halo/galaxy population. The MIP can be used to study statistics of halo properties in cosmic environments where conventional simulations fail such as the interior or voids and walls.

Read the paper
Feb
2007

The spine of the cosmic web.

The Spine is a topological method to classify the cosmic web into its basic morphological components. It is based on a very neat property of the cosmic web: local geometry can be derived from topology. The spine method computes the critical points in the density field and relates their connectivity to a specific geometry. It can find voids (via the watershed method) and from that walls and filaments.

Read the paper
Feb
2007

Hierarchical Multiscale Morphology Filter.

The idea of scale spaces can be generalized (in cosmology at least) to hierarchical spaces where scale of interest is defined at the initial conditions (i.e. linear regime) and then let evolve to the non-linear regime. This basically means that by smoothing the initial conditions we can generate a much cleaner scale space, which we call hierarchical space. This has the advantage of producing a scale space with well defined features similar to anisotropic filtering but in this case made by gravity.

Read the paper