UnderfittingIt refers to the situation where the model has a poor fit to the training data and is usually used to evaluate the model's learning and generalization capabilities.
There are three situations for training data and evaluating models: underfitting, fitting, and overfitting
The impact of underfitting
Underfitting usually occurs in simple models, which cannot reflect the data relationship, resulting in the machine learning model being unable to obtain a low training error. In addition, it may also be caused by factors such as too few feature sets, too few data sets, and unreasonable sampling data.
Solutions to Underfitting
- Find the optimal weight initialization scheme;
- Use appropriate activation functions;
- Choose an appropriate optimizer and learning rate.
Related words: fitting, overfitting
References
【1】https://docs.aws.amazon.com/zh_cn/machine-learning/latest/dg/model-fit-underfitting-vs-overfitting.html
【2】https://blog.ailemon.me/2018/04/09/deep-learning-the-ways-to-solve-underfitting/