CPLUOS Astrophysics team meeting

Friday 7 Oct 2022, 14:00 → 15:20 Asia/Seoul

물리학과

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

COSMOLOGY¶

1 Hyeonmo's Report¶

Speaker: Hyeonmo Koo (University of Seoul)

Introduction of GALIC code

• Use DICE code.

2 Young's report. FoF and MST algorithm¶

Speaker: Young Ju

FoF and MST Algorithm

• Made Mulguisin repository.
• In last meeting, one of the topological  information MGS is weird.
• This time, only plot 50 largest cluster with polar angle.
• There is another variable that is little bit changed form 'Polar angle'.
• In the last meeting, one of the topological information from MGS is weird.

3 Hannah's Report¶

Speaker: Hannah Jhee (University of Seoul)

Halos infalling a cluster through filaments

• No progress

4 Seyeon's Report - CLML¶

Speaker: Se Yeon Hwang (Universe of Seoul)

Predicting Cosmological Parameters Using CNN and ViT from Large Scale Simulation

• Large scale structure
  • Cosmological parameter → Simulation → Output (Large scale structure) → Methodology (2 pt, Machine learning ...)
• Use PINOCCHIO code
  • Ω_m, σ₈, w₀, w_a, n_s, h
• Data
  • Box size = 2 Gpc,
  • Number of particles = 1024³
  • Light cone data
  • Made 3-d histogram.
  • Bin size = 64
• Algorithm
  • CNN (Convolution Neural Newwork)
  • Vit (Vision Transformer)
  • Cvit (Convolution Neural Newwork+Vision Transformer)
• Result with argumentation

5 Sumi's report¶

Speaker: Sumi Kim (University of Seoul)

Cosmology of High-Order Statistics (CHOA)

• Calculation from LSS
• N-point correlation function
  • 1. N-point result
  • 2. Regressor
  • 3. Estimation
  • 4. Contour
    • Goal: Showing the higher order statistics gains more information than the lower ones.

6 John Suarez's report¶

Physical features from Photo+Spect with DL

• Spectroscopic data models
  • z < 0.5
  • (Model, Learning rate, Batch size, Epochs, r²)
    • (resnet50, 0.1, 16, 300, 0.849)
  • p = 0.3
    • Dropout before last linear layer
      • Weights from first 6 layer freezed
    • Resnet50 (6 first layers freezed+3 linearlayers [in_features, 512, 256, 1] → Redshift, log_luminosity
• Deep galaxy net
  • Photometric+spectroscopic data models
    • (Model, Learning rate, Batch size, Epochs, r²)
      • (resnet50, 0.1, 16, 300, 0.909)
• Contrasting with a machine learning method.

7 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

8 Discussion¶