GalaxyHunters / Evolution-Machine-Satellite-Galaxies

Creating a Machine-Learning Model for Prediction of Satellite Galaxy Evolution
0 stars 0 forks source link

Feature Searching #72

Closed Beilinson closed 3 years ago

Beilinson commented 4 years ago

DOD 1 - We need at least 1 more feature to separate the classes. Find it, through whatever means necessary.

Explanation: Many of the overlapping 4 galaxies are actually 3. Solution: Search KDEPLOTS for separation between 1 and 3.

Beilinson commented 4 years ago

FEATURE SEPARATION FOR 1 | 3

Reg:

From sgal_features (sgal_qlabel_0 + 1)

stars_AM*Vorbit(0.9coldgas) sgal_qlabel_0 |
gas_q(0.5kpc) sgal_qlabel_1 |
EllipsDepth_gas(0.5kpc) peri_Vorbit sgal_qlabel_1 |

From (sgal_qlabel_4+3+2)

cen_gas_q(10kpc) |
gas_q(1kpc) | s

Normalized:

From sgal_features (sgal_qlabel_0 + 1)

Rorbit (slight) | center_Msat(0.1rvir)[Msun] / center_Rvir[kpc] |
Rorbit(peri_1/R_vir[kpc] |
density(R0.5kpc)[Msun/kpc^3] (slight) |
zana_r |
stars_q(Rsat) |
EllipsArea_gas(0.9coldgas)_peri_Vorbit |
EllipsArea_stars(Rsat)_peri_Vorbit |
EllipsArea_gas(0.5kpc)_rvir_VorbitSquared |

From (sgal_qlabel_4+3+2)

SFR(R_disc)[Msun/yr] |
density(Rcoldass_0.9)[Msun/kpc^3] |
Mdm(rvir)[Msun] |
Mcold_gas(0.15rvir)[Msun] |

From sgal_forces

gas_phase_metallicity(disc)[log(O/H)+l2] |
Fgrav |
Pgrav |
Ptidal |