Photometric redshift estimation via deep learning - Generalized and pre-classification-less, image based, fully probabilistic redshifts

Free Access

Table 3

Results of Experiment 1 (galaxies).

Criterion	Model	Bias(Δz)	σ²(Δz)	Mad(Δz)	rmse(Δz)	Bias(Δz_norm)	σ²(Δz_norm)	Mad(Δz_norm)	rmse(Δz_norm)

	RF	0.0001	0.0033	0.0164	0.0575	–0.0017	0.0018	0.0133	0.0431
Mean	MDN	0.0016	0.0034	0.0174	0.0589	–0.0003	0.0019	0.0141	0.0442
	DCMDN	0.0018	0.0030	0.0157	0.0548	–0.0003	0.0017	0.0128	0.0409

	RF	0.0031	0.0042	0.0171	0.0652	0.0010	0.0022	0.0139	0.0471
First mode	MDN	0.0029	0.0039	0.0172	0.0628	0.0010	0.0021	0.0140	0.0459
	DCMDN	0.0060	0.0031	0.0167	0.0561	0.0029	0.0016	0.0135	0.0407

	RF	0.0001	0.0023	0.0147	0.0484	–0.0011	0.0013	0.0121	0.0365
Selected	MDN	0.0021	0.0027	0.0165	0.0523	0.0006	0.0015	0.0136	0.0391
	DCMDN	0.0017	0.0023	0.0146	0.0485	–0.0001	0.0013	0.0120	0.0366

Notes. Based on the estimated PDFs the traditional scores have been calculated. This was done by using the mean, the most dominant mode and the mean of the selected unambiguous objects, respectively.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.