I am trying to identify LSB galaxy using unsupervised ML.
- Email of the Prof Claudia
Oi Luís, aqui uma das tabelas (Fornax.csv)
é a que contem 120 mil objetos que tem os
parâmetros corretos para
serem LSBs nos 23 campos de Fornax.
E o outro ConfimedLSBs... são 122 galáxias
LSBs confirmadas da literatura que
junto com as 199 (total de 321) que eu já
tinha enviado são as que vamos usar para treinar.
Uma coisa para diminuir os 120 mil objetos seria
restringir um pouco mais os parâmetros plotando
os Mu_max, Kron_radius
e r_auto de todas as 321 LSBs e so pegar dos 120
mil objetos os que tiverem os parâmetros dentro
do range de parâmetro destas 321.
- I first plotted the parameters:
MU MAX,Kron Radiusandr_auto. The results are in the notebook: Analizyn-LSG.
I then implemented HDBSCAN to the sample of LSB galaxies candidates using as input
parameter MU MAX, Kron Radius and r_auto. It is possible to see the procedure here.
HDBSCAN found 7 cluster:
cluster-hdbscan/LSGs-candidate-HDBSCAN-cluster-0.csv cluster-hdbscan/LSGs-candidate-HDBSCAN-cluster-1.csv cluster-hdbscan/LSGs-candidate-HDBSCAN-cluster-2.csv cluster-hdbscan/LSGs-candidate-HDBSCAN-cluster-3.csv cluster-hdbscan/LSGs-candidate-HDBSCAN-cluster-4.csv cluster-hdbscan/LSGs-candidate-HDBSCAN-cluster-5.csv cluster-hdbscan/LSGs-candidate-HDBSCAN-cluster-6.csv
and the noise:
cluster-hdbscan/LSGs-candidate-HDBSCAN-cluster-noise.csv
- The
cluster-hdbscan/LSGs-candidate-HDBSCAN-cluster-6.csvfile contains the cluster that coincides with confirmed LSB galaxies in the different plots. This cluster has around 80 mil galaxies.