We’ve developed a simple meta-learning algorithm called Reptile which works by repeatedly sampling a task, performing stochastic gradient descent on it, and updating the initial parameters towards the final parameters learned on that task. This method performs as well as MAML, a broadly applicable meta-learning algorithm, while being simpler to implement and more computationally efficient.
Source: openai.com
You may be familiar with matryoshka dolls: nested sets of painted figurines that fit within painted figurines. In the case of wooden dolls, the concept is pretty straightforward: hollow out a large bit and fit smaller bits in. You might think that doing the same thing with atoms is kind of tough.
Read More
Learning has an evolutionary purpose: Among species, individuals that adapt to their environments will succeed. That’s why your brain more easily retains important or surprising information: It takes very little effort to remember that the neighbor’s dog likes to bite. Remembering the dog’s name is harder.
Read More