Question-51: In Machine Learning, why generalization is important?
Answer: In Machine Learning Generalization is important because the data we collected to build and test our model is only sample data and those does not cover all the use cases or possibility in the data. And most of this data is noisy as well.
Question-52: What is the problem, if you over train your Machine Learning Model?
Answer: When you overtrain your Machine Learning Model on the training data set then certainly it would identify all the relevant information in the training data. However, it would most likely fail to make correct prediction when you input completely new data to your model. And this is the situation where you can say that model is incapable of generalizing or that it is overfitting the training data. And also, you train your model more and more, it will become more complex as well.
Question-53: What is the problem with the over trained model?
Answer: As we have seen in previous Question, as you train more and more your model. Its complexity increases and, on the training, data sets its error reduces. And you think your model would perform better. However, as soon as you present new data to this model it would fail in prediction.
Question-54: What do you mean by Machine Learning Model Generalization?
Answer: Whatever, model you are creating it should be able to Generalize well. It means, you should not overtrain your Model with the training data set. Hence, to predict a good Advanced Machine Learning Model it should be capable of generalizing, and you should know when to stop training the model and it does not overfit. Best example in layman term, don’t teach each and every detail to your child. Let them learn something from real world. They should be able to generalize new situations based on the learning/teaching you taught them. Otherwise, they would always depend on parent and when parent is not there. They would miserably fail to perform in new situations or conditions.
Question-55: How data quality affects the Model Generalization?
Answer: Hence, to have model well generalized, you must ensure that the data used to train model is good and reliable sample of observations. The higher the quality of the data and more representative the sample is the easier it will be for model to learn the unknow mapping.
Generalization, means you predict well on the unseen data.