From the course: LinkedIn AI Academy AI-100: 1 Demystifying AI
Join today to access over 23,200 courses taught by industry experts.
Bias and variance
From the course: LinkedIn AI Academy AI-100: 1 Demystifying AI
Bias and variance
- [Instructor] In multiple linear regression, we saw that there are many possibilities in the ways we can model the predicted variable using the features. But how do we find out which one of these possibilities is the best method for our model? This is where we come up with two very important concepts in machine learning. Bias is a measure of how accurately the model can learn at best. If you are using a simple linear model on a dataset where the relationship is more complex, the model might not predict the data very accurately and will end up with a high bias. When a model is complex, it isn't just linear, but quadric or cubic in features, for instance, that tends to give the model more capacity to learn and be more accurate. So we say the model has a lower bias. This seems great, but there's a trade off. The model might have a low bias, but it's not as easy to train. This is where variance comes in. Variance…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.