[liangqin12354]

Hi,

I just found there might be a error in the provided regression score function.

It seems you don’t use the new area in the score calculation?

Best regards,
Qin

---

Extract the scipy model

model = model_PDFs[pdf_type]

The following block of code generates the full PDF for potential normalization

Generate x values for plotting and integration

n_points = 100000
x = np.linspace(-100, 100, n_points)
y = model.pdf(x, **pdf_args)

Verify the area under the curve using numerical integration

area = simps(y, x)

Normalize the PDF if the area is not 1

if area != 1:
y /= area
area = simps(y, x)

# Calculate the score with corrected area
score = model.pdf(true_target, **pdf_args)

Normalize score if the the PDF max is greater than 1

y_max = y.max()
if y_max > 1:
y = y/y_max

score = score/y_max

[thejbills]

GOOD catch. We’re updating that.

[Author]

Posts