Explain the graph error histogram with 20 bins in neural network. what is bins?

How can we explain the graph error histogram with 20 bins in neural network. what are bins?

Answer 1

[Ameer Hamza]

Bins are the number of vertical bars you are observing on the graph. The total error from neural network ranges from -0.2222 (leftmost bin) to 0.1968 (rightmost bin). This error range is divided into 20 smaller bins, so each bin has a width of

(0.1968-(-0.2222))/20 = 0.02095.

Each vertical bar represents the number of samples from your dataset, which lies in a particular bin. For example, at the mid of your graph, you have a bin corresponding to the error of -0.00166 and the height of that bin for validation dataset is 10. It means that 10 samples from you validation dataset have an error lies in the following range.

(-0.00166 - 0.02095/2 , -0.00166 + 0.02095/2)
(-0.012135, 0.008815) < the range of the bin corrosponding to -0.00166

Similarly, for other bins, you can interpret the result.

[NEHA SINGH]

Why there is no test result in the middle of the bin?

[Ameer Hamza]

It is because there is no sample in you test dataset which gives an error in this range. If you manually find the error for your test dataset, you will observe that there is no error lying in this bin.

[NEHA SINGH]

Thank you Sir Request you to explain the attached graph also.

[Ameer Hamza]

This shows how accurately your trained model fits the dataset. If the value of R-squared is close to 1 (good) then it shows that the model prediction is very close to the actual dataset. If it is zero (bad) then it shows that you model completely fails in making a correct prediction. You can get more information from here:

http://blog.minitab.com/blog/adventures-in-statistics-2/regression-analysis-how-do-i-interpret-r-squared-and-assess-the-goodness-of-fit

https://en.wikipedia.org/wiki/Coefficient_of_determination

[NEHA SINGH]

Thank you sir.

[Yu Hsiang Lin]

Are the errors in terms of prediction minus real data? or are the errors some factor of mean square error?