optimization - Learning Curves for Multi-Class Logistic Regression -
i have written multi-class classifier using logistic regression trained using 1 vs approach. want plot learning curve trained classifier.
should learning curve plotted on class class basis or should single plot classifier whole? make difference?
to clarify, learning curve plot of training & cross validation/test set error/cost vs training set size. plot should allow see if increases training set size improves performance. more generally, learning curve allows identify whether algorithm suffers bias (under fitting) or variance (over fitting) problem.
some details regarding code:
- analyzes mnist handwritten digit images
- predicts digit (0-9) in image
- based on andrew ng's coursera class on machine learning
generally, i'd plot things. or write script collect , plot all.
i think need whole classifier plot obvious. class-by-class ones strike me valuable ensure you're not having problems single class. if, say, "5" stubbornly resistant increased training data, overall classifier still being helped it, i'd prefer investigate situation 1 class before poured on more data.
Comments
Post a Comment