Why log scale is used in the function learn.predict

Part 1 (2018)
1

This is from video 1 and looking for clarifications on following two questions

log_preds = learn.predict(). Why are the prediction being returned in log scale? The predictions are already probabilities. why do we need to take it to log scale?

What is the significance of negative logarithmic predictions
array([[ -0.00002, -11.07446],
[ -0.00138, -6.58385],
[ -0.00083, -7.09025],
[ -0.00029, -8.13645],
[ -0.00035, -7.9663 ],
[ -0.00029, -8.15125],
[ -0.00002, -10.82139],
[ -0.00003, -10.33846],
[ -0.00323, -5.73731],
[ -0.0001 , -9.21326]]

Thank you in advance

1 Like
2

when you multiply small numbers you get even small numbers. if you do it repeatedly, within few steps the final value vanishes. To overcome this, small numbers are expresses in logarithm scale.

Please note, multiplications become additions in log scale.

6 Likes
3

Thank you

4

Can you explain a bit more? Are we applying log on the final output given by the predict function?