Prediction accuracy bugged

[sjg2203]

I am trying to compare both manual and predicted scoring using this:

hypno_pred=sls.predict()
hypno_manual=np.loadtxt(easygui.fileopenbox(msg="Annotation file"),dtype=str)
accuracy=(hypno_manual==hypno_pred).sum()/hypno_pred.size
print("The overall agreement is %.3f"%accuracy)

However, this is always what comes out of it:

accuracy=(hypno_manual==hypno_pred).sum()/hypno_pred.size
AttributeError: 'bool' object has no attribute 'sum'

Here

yasa/notebooks/14_automatic_sleep_staging.ipynb

Line 228 in b98a4c0

"The overall agreement is 0.837\n"

(box nº6), it's showed as working but when reproducing the same thing, an error is prompted.
Does anyone know why?

Thanks

[remrama]

It might be a typo where there are missing parentheses on sls.predict(). The first line in your example code should be:

hypno_pred = sls.predict()

If that doesn't solve it, it might have something to do with the easygui part. If your problem isn't solved by adding parentheses to sls.predict(), please show the output from:

print(hypno_manual)

[raphaelvallat]

Can you try with: hypno_manual = np.array(hypno_manual)?

[sjg2203]

hypno_manual is already an array, though I tried yours but did not work

[raphaelvallat]

Can you print(hypno_pred) also please?

[remrama]

@sjg2203 the problem is in this section: hypno_manual==hypno_pred

Whereas we would expect hypno_manual==hypno_pred to result in a numpy array full of booleans (Trues and Falses), but for you it's a single boolean. To get the behavior we are expecting, it's important that both hypno_manual and hypno_pred are numpy arrays of the same length.

Double-check that they are both numpy arrays. If the problem persists, please share your output after running the following:

hypno_pred=sls.predict()
hypno_manual=np.loadtxt(easygui.fileopenbox(msg="Annotation file"),dtype=str)

hypno_pred = np.asarray(hypno_pred)
hypno_manual = np.asarray(hypno_manual)

print("Numpy version:", np.__version__)
print("Hypno pred type:", type(hypno_pred))
print("Hypno pred dtype:", hypno_pred.dtype)
print("Hypno pred shape:", hypno_pred.shape)
print("Hypno manual type:", type(hypno_manual))
print("Hypno manual dtype:", hypno_manual.dtype)
print("Hypno manual shape:", hypno_manual.shape)

agreement = (hypno_manual == hypno_pred)

print("Agreement:", agreement)
print("Agreement type:", type(agreement))
print("Agreement dtype:", agreement.dtype)
print("Agreement shape:", agreement.shape)

[sjg2203]

My var "hypno_manual" was a bit shorter than the other var, solved the problem by upscaling it, works great now!
Thanks