In Machine learning, learning rate is a hyper parameter in gradient descent algorithm that decides the step size for each iteration while moving towards the convergence point. The convergence point is nothing but a condition when to stop the algorithm. We should choose a learning rate such that there won’t be any overshooting.
- If the learning rate is too small, the algorithm takes a lot of time to reach the convergence point
- If the learning rate is too high, the algorithm keep on jumping over the convergence and may not meet convergence point at all.
So, we need to choose a learning rate considering both conditions.
You can register for this Machine learning certification program by Intellipaat to learn Machine learning.
You can watch this video on Machine learning Full course to know more about learning rate and how to tune the learning rate: