Stochastic Hopfield net Boltzmann machine is nothing but stochastic Hopfield net1. If you did not yet read the post of the Hopfield net in the blog, just go read it. I assume the readers are familiar to it, and directly use many results we had in the post. The magic of deep learning which we have discussed a couple of times works here, too. Such as $\epsilon$-greedy off-policy algorithm2, the stochastic character of the binary units allows the machine occasionally increase its energy to escape from poor local minima.

Continue reading

Judgement Day It is the first time I did not post for 4 days. I was too busy to prepare for the meetup this week. The day before yesterday meetup topic was the reinforcement learning as I mentioned at previous post. It is not a long research paper, but includes 143 references. Ah, not my favorite. This A Brief Survey of Deep Reinforcement Learning did not explain the detail of what I am interested in.

Continue reading

Hard K-means and responsibilities If you did not read the first part of the clustering series. Please go check it out. I use the same data points and this post starts from troubleshooting the hard K-means algorithm in the previous post. In the previous post, we defined assignment. The equivalent representation of this assignment of points to clusters is given by responsibilities, $r^{(n)}_k$. In the assignment step, we set $r^{(n)}_k$ to one if mean k is the closest mean to datapoint $ {\textbf x}^{(n)}$; otherwise, $r^{(n)}_k$ is zero.

Continue reading