CPLUOS Astrophysics team meeting

Friday 17 Feb 2023, 14:00 → 15:20 Asia/Seoul

물리학과

https://uos-ac-kr.zoom.us/j/8264902652

14:00 → 15:00 COSMOLOGY

14:00 Hyeonmo's Report

Speaker: Hyeonmo Koo (University of Seoul)

Planar System Generation

• Prepare KPS presentation.

14:10 Young's report. FoF and MST algorithm

Speaker: Young Ju

FoF and MST algorithm

• We checked what Dr. Hong said and took a note.
• Test free parameters for DBSCAN.
  • min_sample-number of clusters (log scale) graph
• Future plan
  • Make more topological parameters.
  • Constrain cosmological parameters using MGS clusters.
    • Using Graph Neural Network (GNN) and non-parametric regression.
      • Graph Neural Network
        • FoF
        • MST
        • DBSCAN
        • MGS

14:20 Hannah's Report

Speaker: Hannah Jhee (University of Seoul)

Active Galactic Nucleus (AGN)-Merger Connectivity

• Generating Mock Images
  • Pristine
  • PSF+Noise
  • Full synthetic image

14:30 Seyeon's Report - CLML

Speaker: Se Yeon Hwang (Universe of Seoul)

Cosmology with Large scale structure using Machine Learning (CLML)

• Pearson correlation  coefficient: Measure of linear correlation between two sets of data.
  • ρ_{X, Y} = cov (X, Y)/(ρ_Xρ_Y)
    • ρ = -1
    • -1 < ρ < 0
    • ρ = 0
    • 0 < ρ < 1
    • ρ = 1
• Our goal: See correlation between CNN, ViT and 2pcf result.
• σ₈ and w₀ have lower correlation between them.
• Linear correlation process
  • 1. Make range.
  • 2. Calculate mean inside range (blue point).
  • 3. Fitting blue points, what we to do is moving yellow scatter.
  • 4. Move points.
• Custom loss function
• Check activation function.
• Note tangent hyperbolic.
• Collect data as weight.

14:40 Sumi's report

Speaker: Sumi Kim (University of Seoul)

Cosmology of High-Order Statistics (CHOA) & Dark Energy Spectroscopic Instrument (DESI)

• Previously on CHOA
• Methodology
  • Paired fixed field
  • Same power spectrum with standard  Gaussian fields sample variance and reduce scatter.
  • But without linear scale amplitude scatter
  • Can suppress.
  • Refer power spectrum k.
• Simulation
  • Used different random seed for gaussian.
  • Used same  random seed for pff pair.
  • Both errors calculated from pff pair(s) mean are well distributed in 1-sigma range.
  • Used halo particles in calculating N_pcf.
    • Similar feature with galaxy
  • Now generating simulations for paired fixed field.
• Separately
  • f_R(R) change: Probing scalar tensor theories for gravity in redshift  space.
  • Brans-Dicke: Brans-Dicke theory.

14:50 Dr.Sabiu's Report

Speaker: Dr Cristiano Sabiu (University of Seoul)

Deep Learning the Deep Sky: Recovering low surface brightness objects with machine learning

• No progress

15:00 → 15:10 Discussion