Why does logistic cost curve continues to decline, when the cost is small?

In Machine Learning Specialization in the Lab “Logistic loss” it was said that when the cost is small, the logistic cost curve has a slope and continues to decline. I just can’t get it. How can it continue to decline if it has the global minimum? Or is it meant that for these values(visible on the plot) it continues to decline?
Thanks in advance!:slight_smile:

1 Like

The “AI Questions” forum is for general topics. Not individual course topics.

Please move your thread to the correct forum area for your course.

Each course has a forum in the “Course Q&A” area.

You can move your thread by using the “pencil” icon in the thread title.

The cost continues to decrease if and only if you haven’t hit the minimum yet. Of course it depends on how sophisticated your gradient descent algorithm actually is: if you are using a fixed learning rate, you may well overshoot the minimum and not have a monotonically decreasing cost.

Of course as Tom says, someone may be able to give a more complete or more specific answer if you give us the context of which lecture in MLS you are referring to here. The time offset would be helpful.

ok, thank you, moved it to “Course Q&A”

1 Like

Thank you:)