Monday, August 1, 2016

Deep AI Making Strides

Inceptionism Iteration

Born from my dual interest in both building systems and neural networks, this post is a bit off-track for Limbic Signal, but not really – we’re looking here at neural networks, the ones mentioned in Hidden Scents. The neural networks used to run Google’s Deep Mind are similar to the workings of the olfactory bulb in the way they use layers of feedback systems to recognize patterns.

Deep Mind is in the news because it cut the electricity bills at one of Google’s buildings by a lot. The thing about these self-learning algorithms, if you will, is that the way they work, or how they work, is really unknown to us. They are using an optimization algorithm to generate their results, which means making microtweaks on hundreds of variables and in realtime. It’s the opposite of a silver bullet approach to energy efficiency (and it's also the way we learn to smell, and why smells mean different things to different people). I’ll let the researchers themselves talk about it; this is from their blog:

20 JULY 2016, Rich Evans, Research Engineer, DeepMind and Jim Gao, Data Centre Engineer, Google

“Each data centre has a unique architecture and environment. A custom-tuned model for one system may not be applicable to another. Therefore, a general intelligence framework is needed to understand the data centre’s interactions.

“We accomplished this by taking the historical data that had already been collected by thousands of sensors within the data centre -- data such as temperatures, power, pump speeds, setpoints, etc. -- and using it to train an ensemble of deep neural networks. Since our objective was to improve data centre energy efficiency, we trained the neural networks on the average future PUE (Power Usage Effectiveness), which is defined as the ratio of the total building energy usage to the IT energy usage. We then trained two additional ensembles of deep neural networks to predict the future temperature and pressure of the data centre over the next hour. The purpose of these predictions is to simulate the recommended actions from the PUE model, to ensure that we do not go beyond any operating constraints.

“Our machine learning system was able to consistently achieve a 40 percent reduction in the amount of energy used for cooling, which equates to a 15 percent reduction in overall PUE overhead after accounting for electrical losses and other non-cooling inefficiencies. It also produced the lowest PUE the site had ever seen.


And furthermore, I’ve taken a piece from another one of their posts:

17TH JUNE 2016, David Silver, Google DeepMind

“However, deep Q-networks are only one way to solve the deep RL problem. We recently introduced an even more practical and effective method based on asynchronous RL. This approach exploits the multithreading capabilities of standard CPUs. The idea is to execute many instances of our agent in parallel, but using a shared model. This provides a viable alternative to experience replay, since parallelisation also diversifies and decorrelates the data. Our asynchronous actor-critic algorithm, A3C, combines a deep Q-network with a deep policy network for selecting actions. It achieves state-of-the-art results, using a fraction of the training time of DQN and a fraction of the resource consumption of Gorila. By building novel approaches to intrinsic motivation andtemporally abstract planning, we have also achieved breakthrough results in the most notoriously challenging Atari games, such as Montezuma’s Revenge.”

Post Script
I can’t talk about Deep Mind without mentioning Deep Dream though: check out what it looks like for a computer to dream, it’s basically a new artform called Inceptionism, and it comes from these neural networks.

No comments:

Post a Comment