Learn this before starting with machine learning
You are on a very good start if you learn Backpropagation and use it to play with some Neural Networks and then learn Restricted Boltzmann Machines and contrastive divergence.
The path I would recommend is:
- Linear Regression and Gradient Descent
- Logistic Regression
- Neural Networks and Backpropagation
- Autoencoders
- Restricted Boltzmann Machines and Contrastive Divergence
- Deep Belief Networkd
- Convolutional Networks
And you are ready to try deep architectures.
I strongly advise you to actually code a neural network with backpropagation, a RBM with contrastive divergence and an autoencoder if it's easy to find quick applications for those to get you interested.
I don't know what kind of programming language you use but I'd recommend you to use Python,Matlab/Octave or R. If you know Python Theano will be a very good way to experiment with deep networks.