Metrics and loss

These functions are taken mostly from sakitools, which is developed by Fabian hickert

---

source

dice_coef

 dice_coef (y_true, y_pred, smooth=0)

---

source

tversky_focal_loss

 tversky_focal_loss (alpha=0.5, beta=0.5, gamma=1.5, name=None)

---

source

dice_loss

 dice_loss (y_true, y_pred)

---

source

dice_coef_loss

 dice_coef_loss (y_true, y_pred)

---

source

FDR

 FDR (thresholds=0.5, name='fdr_overreject', **kwargs)

! FDR - False discovery rate metric FP / (FP + TP)

---

source

FNR

 FNR (thresholds=0.5, name='fnr_escapee', **kwargs)

! FNR - False negative rate metric FN / (FN + TP)

---

source

tversky_index

 tversky_index (alpha=0.5, beta=0.5, gamma=1.5, name=None)