"Sparse GPs: approximate the posterior, not the model" by @jameshensman https://t.co/XG8E2t0h6S #ai #MachineLearning read more
under the hood
The burgeoning AI industry is still focused on perception – recognising images, understanding speech, helping robots touch. But far greater benefits will come when AIs can make good, principled decisions based on those perceptions.
PROWLER.io’s Decision Making systems:
learn from experience
use Probabilistic models that are flexible enough to generalise to novel situations
adapt to changing environments by refining their strategies in data-driven ways
continually update their environmental models in light of data collected while operating.
use Reinforcement Learning (RL) algorithms that can help estimate, account for and even reduce uncertainty
can cope with the uncertainties of finite – even small – amounts of data
can favour decisions that loop back fresh information to models
use Game Theory to operate in multi-agent settings
can cooperate – and compete – by inferring what humans and other AIs are trying to do.