A review of how to evaluate probabilities from densities over continuous variables. Then, a discussion and implementation of how to apply these ideas within latent variable models.
A discussion of intelligence, predictive coding, and machine learning.
A discussion and tutorial on statistical whitening.
Thoughts about the importance of data for machine learning and humans.
Generative probabilistic models, combined with deep neural networks, demonstrate the possiblilty of extracting meaningful abstract representations in unsupervised ways.
In this blog post, I'll show you how I implemented GoogLeNet in Keras and copied over the weights from Caffe. Then we'll classify some cats!
Backpropagation is a method for computing derivatives in artificial neural networks, allowing us to use gradient descent to train these models. Here, I walk through implementing backpropagation.