An Introduction to Variational Autoencoders

Variational Autoencoders (VAE) are really cool machine learning models that can generate new data. It means a VAE trained on thousands of human faces can new human faces as shown above!

Recently, two types of generative models have been popular in the machine learning community, namely, Generative Adversarial Networks (GAN) and VAEs. While GANs have had more success so far, a recent Deepmind paper showed that VAEs can yield results competitive to state of the art GAN models. Furthermore, VAE generated images retain more of the diversity of training dataset than GAN counterparts. Continue reading An Introduction to Variational Autoencoders

The Maths Behind Logistic Regression

Let’s have a look at this god-tier math puzzle. Only one out of seven gets it right, and the other six don’t. So, what are the odds for solving it correctly? 1 to 6. Generally, if $$p$$ is the probability that someone will get it right, then his/her odds are $$p/(1-p)$$.
However, it isn’t necessarily true that $$p=1/7$$ for every person because some people are smarter, some have better education and so on. Hence, $$p$$ also depends on the person attempting the puzzle. In a Bayesian framework, we capture this dependence with conditional probability.