Making better decisions demands a toolbox, not a single tool.
Our technology explained
In business, you need to make certain types of decisions on a regular basis. Perhaps setting prices in markets with volatile demand. Managing inventory and optimising stock levels. Or even predicting the impact of ‘unpredictable’ events, like the weather or traffic jams.
To speed up these decisions, we've developed a unique ecosystem that can tailor probabilistic models and decision-making algorithms to your business challenges.
Our foundation: Gaussian Process
We make widespread use of Gaussian Processes, or “GPs.” Compared to the more fashionable Deep Neural Network machine learning approach, GPs have a number of advantages:
Probabilistic modelling libraries
Our Probabilistic modelling libraries account for all the scenarios that are compatible with both expert knowledge and available data. This is particularly data efficient since the specification of expert knowledge already provides a lot of information to the model. As a result, you can make real-time decisions with a limited amount of data, reducing computational costs, data and energy requirements.
The environment in which we all act is complex and dynamic. So, we need adaptive solutions that can work out the best course of action given a model’s prediction, taking into account the dynamic relationship between the model and the environment that we’re acting in. We may also need to consider what actions (and associated outcomes) are going to be available to us in the future as a result of the immediate next action. Working out optimal action is a hard problem, that's where our Decision libraries come in.
AutoGP takes your data and finds the best suited probabilistic model for it. This is done by refining the model in a smart way that limits the number of models to be tested. Our AutoGP is uniquely powerful, because:
Partners and integrations
Interested in learning more about becoming a partner or how to easily integrate with the Secondmind Decision Engine?