In 5:11,andrew said that In the past, people were using linear methods, now they’re using convolutional neural networks, so was the old method the fully connected method that you’ll see in the next few videos? So now we’re using convolutional neural networks, right?
Hey @boe,
At 4:50, Prof Andrew says " So, before the rise of Neural Networks people used to use much simpler classifiers like a simple linear classifier over hand engineer features in order to perform object detection."
In my opinion, here, Prof Andrew is referring to simple machine learning models such as Logistic Regression. This is because fully connected neural networks with non-linear activations such as ReLU, Sigmoid, etc, won’t be referred to as “linear classifiers”.
Now, you may wonder that even Logistic Regression involves the use of the sigmoid function to determine the decision boundary, so how can we refer to it as a linear model? For that, you can find a great many resources on the web, for instance, you can check this thread out. For a more detailed explanation, check this amazing blog out. I hope this helps.
Cheers,
Elemento