OCEANS OF NOISE

TECHNOLOGY

-

VISUAL ART

-

MATHEMATICS

-

TECHNOLOGY - VISUAL ART - MATHEMATICS -

Oceans of Noise 1/2

Hyper Suprime-Cam (HSC) Year 1

Synthetic covariance matrices for photometric  redshift estimation of galaxies

By Jaime Ruiz Zapatero (2024) - Covariant Art 

... Identifying the relative positions of galaxies is the most powerful way we have, to learn about the impact of gravity in our Universe historically and today.
— Jaime

Oceans of Noise is a collection of pieces by Dr Jaime Ruiz Zapatero that show mathematical representations of errors from a huge (22.2 m) telescope in Hawaii used to observe galaxies. Regardless of such a powerful tool, challenges to its accuracy still exist to determine the exact distances of other galaxies, mainly due to techniques used in this process that e.g., are limited to a few colour filters. However, Jaime expressed that ‘future better light detectors will allow much better pictures’ which could reduce these errors.

Oceans of Noise 2/2

Hyper Suprime-Cam (HSC) Year 1

Synthetic covariance matrices for photometric  redshift estimation of galaxies

By Jaime Ruiz Zapatero (2024) - Covariant Art 

The illuminated patterns in Oceans of Noise remind us of Irma de Vries’ collection of installations at the Moco Museum (Barcelona and Amsterdam), including Kalidoscope, Connect the Dots & Universe, and Diamond Matrix. These are immersive art experiences in spaces consisting of augmented reality, sculpture and lights. These pieces highlight the concept that whilst the entire galaxy is made of the same particle matter, every individual one has connections to others crucial in creating the bigger system that surrounds and is us.

Kaleidoscope by Irma de Vires

Image credits: @studioirma_art

Kaleidoscope emphasises the importance of widening one’s scope, beyond what immediately seems like repeating patterns, to see different arrangements and meanings emerging. This concept aligns with Jaime’s piece, with the idea that we can visualise differences in patterns of the matrix, which encourages a curiosity for its implications in the context of our understanding of the telescope’s observations.

...figuring out how far away other galaxies are from our own remains a surprisingly challenging task and a major source of uncertainty...
— Jaime

Jaime is a research software engineer at the Advanced Research Computing (ARC) Centre at University College London. He has a wide range of expertise including classical machine learning, high dimensional statistical inference, big data reduction and auto-differentiable programming. His current work is on developing the infrastructure for some cosmological surveys whilst he also has a keen interest in accelerating Bayesian inference with gradient methods and Gaussian processes as tools for model-agnostic science.

A description from Jaime:

Despite our best efforts, figuring out how far away other galaxies are from our own remains a surprisingly challenging task and a major source of uncertainty in modern studies of the large-scale structure of the Universe.

While eye-catching, the strong oscillatory patterns present in these matrices spell doom for our dear theoretical model...
— Jaime

The main source of this uncertainty is the fact that most galaxy observations rely on short-exposure pictures of galaxies through a small selection of colour filters. This observational technique is known as “photometry”. Thus, uncertainties in the positions of galaxies are just a reflection of the limited information on the type of light emitted by said galaxies.

Covariance matrices are a multi-dimensional extension of the concept of error. Hence, they are well suited to represent the uncertainty in the positions of millions of galaxies. The covariance matrices shown here belong to the first-year observations of the Hyper Suprime-Cam, a 900-megapixel digital camera, mounted on the Subaru telescope in Hawaii.

These particular matrices are synthetic, meaning that they have not been directly measured from the data. Instead, they have been generated using a theoretical model for what we expect said errors to be. While eye-catching, the strong oscillatory patterns present in these matrices spell doom for our dear theoretical model as they hint at the model being too simple to capture the error in the actual observations.

We also asked Jaime:

What is the significance behind the title of the piece?

Highly oscillatory noise patterns are captured in covariance matrices that resemble ocean waves.

Welcome JAIME!

I enjoy looking through the plots I might have seen a thousand times at work and wondering if some of them might be eye-catching to somebody else!

 
Soyoung Choi

Founder of STEAMUL8

Next
Next

Sunshine in an AI Black Box