tool g
The situation that the model remains simple enough to learn the images within the data.
Insufficient learning (underfitting) is the case that a model will not be able to capture the basic images in the education data, or because it is insufficient of the educational process, both in the training data and low success in the test data. The opposite problem of excessive learning is: instead of memorizing model data, there is no enough complexity or training time to grasp the relationship within the data. For example, in reality, it is a typical example of learning to work in fitting a linear regression model to a set of data with a curved relationship; the model produces systematically incorrect predictions in both training and testing data.
Among the causes of unfamiliar learning, the capacity of the model (the number of strings, the number of parameters) is low, the short stay of the training period, the insufficient or unacceptable selection of attributes or excessive harsh implementation of regularization. Solution ways are often reversed by these reasons: using a more complex or deep model architecture, prolong training time, produce more meaningful attributes or reduce the regularization power. The practice of machine learning in the app aims to find the right balance between these two ends (bias-variance trade-off) without decreasing what extreme learning is.
Among the causes of unfamiliar learning, the capacity of the model (the number of strings, the number of parameters) is low, the short stay of the training period, the insufficient or unacceptable selection of attributes or excessive harsh implementation of regularization. Solution ways are often reversed by these reasons: using a more complex or deep model architecture, prolong training time, produce more meaningful attributes or reduce the regularization power. The practice of machine learning in the app aims to find the right balance between these two ends (bias-variance trade-off) without decreasing what extreme learning is.