What is the “m” cut-off to use deep learning technique over traditional MLs
In the week-1 tutorial. It is mentioned that for smaller “m” SVM might be better. Since, training the algorithm has a certain cost. How to arrive at the interval of “m” to switch from traditional ML to Deep-Learning.
“SVM” isn’t really a “deep learning” technique. It’s just a machine learning method, and frankly isn’t very popular now because it’s very computationally intensive.
The issue isn’t just the number of examples, it’s the total size of the dataset (examples and features).
Depending on how often you need to train the system again, an SVM might not be practical.
There are no fixed rules for the number of examples. It all depends on what task you’re trying to perform.