Metrics

Comparison of results from DL pickers with ground truth (high-quality picks).

True positives ($t_p$) are picks done by the Deep Learning picker that coincide within a small time difference ($\Delta t < \epsilon$) with the picks from the labelled dataset (e.g. refined IGN picks for El Hierro pre-eruption or CASTOR sequence).

True negatives ($t_n$)

Comparison metrics that we will use are:

$$Precision = { t_p \over {t_p + f_p} } ,$$

$$Recall = { t_p \over {t_p + f_n} } ,$$