Projects at LAM

2019-12-19

08:18 CFHTLS_z Action #5497: CFHTLS training tests with high resolution photo-z added

> +*TRAINING TESTS:*+
>
> model "x" (Jo&Je settings) :
> learning rate = 0.0001 to iteration 150000
> learni... Stephane Arnouts

2019-12-18

18:32 CFHTLS_z Action #5497: CFHTLS training tests with high resolution photo-z added

> The models aren't significantly different but the "u" (and even "v") trainings run in half the time as "x" (~4h f... Stephane Arnouts

2019-02-28

17:32 CFHTLS_z Action #5184: Plots and stats from perfAllDataTest.npy

Marie TREYER wrote:
ssuper. On voit le probleme de taille d'entrainement a bas z et faible magnitude
Stephane Arnouts

16:12 CFHTLS_z Action #5184: Plots and stats from perfAllDataTest.npy

> non ce ne sont pas les mêmes objects. pour le CNN ce sont les 19,500 galaxies traités par Johanna, DEEP and WIDE ... Stephane Arnouts

15:52 CFHTLS_z Action #5184: Plots and stats from perfAllDataTest.npy

Johanna Pasquet wrote:
>
> On voit vraiment bien l'amélioration :) Merci Marie!
oui mais j'ai des doutes sur la ... Stephane Arnouts

15:45 CFHTLS_z Action #5184: Plots and stats from perfAllDataTest.npy

Marie TREYER wrote:
merci Marie. C'est pour les memes objets? Tu pourrais regarder le CNN avec le DEEP aussi et s... Stephane Arnouts

2019-02-14

16:09 CFHTLS_z Action #5009: First results on CFHTLS data

Johanna Pasquet wrote:
> Stephane ARNOUTS wrote:
> > Merci Johanna,
> > ca a l'air tres bon ces resultats !
> ... Stephane Arnouts

12:11 CFHTLS_z Action #5009: First results on CFHTLS data

and we still see the bias problem at high z where Zcnn tend to under-estimate the redshift Stephane Arnouts

12:07 CFHTLS_z Action #5009: First results on CFHTLS data

Merci Johanna,
ca a l'air tres bon ces resultats !
J'ai un doute sur la definition des outliers ?
Il faut le... Stephane Arnouts