CPLUOS Astrophysics team meeting

Monday 13 Sept 2021, 16:30 → 17:55 Asia/Seoul

물리학과

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

16:30 → 17:45 COSMOLOGY¶

16:30 Dr.Sangnam Park's Report - FDM_Offset¶

Speaker: Sangnam Park

• Corotation of Satellites
  • Initial ratio between number of satellites with initial velocity v^+_0 and v^-_0 = 3:1
  • CDM : 0.25 → 0.1934 ± 0.3560
  • FDM : 0.25 → 0.0664 ± 0.0526

16:45 Hyeonmo's report - CDM Halo Evolution for Gadget2 Simulation¶

Speaker: Hyeonmo Koo (University of Seoul)

17:00 Young's report. FoF and MST algorithm¶

Speaker: Young Ju

• Test MGS, MST in Noise data 4
  • Added 3000, 4000 noise galaxies in the random data set.
  • Check the number of cluster
  • Cluster with more than 3 member
    • 3000 noise galaxies ; MST : 273 / MGS : 344
    • 4000 noise galaxies ; MST : 306 / MGS : 344
  • Cluster with more than 10 member
    • 3000 noise galaxies ; MST : 63 / MGS : 63
    • 4000 noise galaxies ; MST : 79 / MGS : 63
  • None noise data
    • MST : 44 / MGS : 47
  • Limit the number of member, the number of cluster is higher than none noise data results.
  • In MGS, 10 clusters and MST, 10 clusters, only pick up largest 10 cluster, there is evident difference.

 

17:15 CHOA(Cosmology of High-Order Statistics)¶

Speaker: Sumi Kim (University of Seoul)

• Compare Abacus Real & Pinocchio with Observations
  • To compare which simulation more fits with data
  • Abacus Parameters
    • Om = 0.3142
    • Ode = 0.6858
    • w = -1
    • Z_b = 0.5
    • Box size = 1100
    • H0 = 67.62
  • In Abacus & Observation 2pcf Comparison,
    • About 100r(Mpc/h) : Observation amplitude is bigger than abacus
    • About 30r(Mpc/h) : Fits with RSD in lowscale

17:30 Dr.Sabiu's Report - Higher Order Statistics¶

Speaker: Dr Cristiano Sabiu (University of Seoul)

• Estimating the General N-point Statistics
  • Query for N-point
    • All 'pairs' (w/ distances) are precomputed.
    • N-point exact counting.
    • I.e. No approximation
    • The majority of the time is taken in binning in high-dimension.
    • N(r_(1-2)) = Fast binning! Combinitronics(we know all the combinations!)
    • N(r_(3-4)) = Slower binning Need to search the database
  • Graph Query (Disconnected Part)
    • To bin the N-point measurement we need the distance between disconnected galaxies (point).
    • Database Entry for Galaxy 265
      • (Distance, Unique ID)
        • (13, 342)
        • (17, 620)
        • (14, 3675) → New method :  Divide-and-Conquer
        • (..., ...)
        • (7, 13465)
    • Generative Pre-trained Transformer 3 (GPT-3) is an autoregressive language model that uses deep learning  to produce human-like text.
    • GPT-3's full version has a capacity of 175 billion machine learning parameters (10x more than any previous model).
    • Back to the N-points : How much faster can we make this?
      • When the number of data points reaches 500000, the speed of new is about twice as fast as the old.

17:45 → 17:55 Discussion¶