Ideal Bais and Varriance achievable


I was curious to know what is the ideal bias and variance that we can achieve ( here I mean, how low it can be in usual deep learning model using larger, neural network, layers, large set of data having lower bayes error) in a used case scenario.

It would be great if someone can share their experience where they have got extremely low bias and variance and if possible please share the techniques you followed to achieve it in specific used case.


In theory, it should be possible to get rid of all avoidable bias and also the variance. In practice, it depends on the task at hand. Machine learning models are better at some tasks than others. Have a look at

to find the best results we can achieve for a specific task.

1 Like