Machine Learning at Scale | More than Deep Neural Networks
04 Apr 2018
Yes
-  

 

 

Machine Learning underpins many of our digital assets, in particular the use of Bayesian Optimisation. 

Yes

​​

 

Abstract colorful background-dreamstime_xxl-Iscatel_19983451 (1900x1267).jpg 

We have developed a range of advanced machine learning capabilities, covering two classes of application:

Bayesian Optimisation
Image detection and classification
​A flexible and efficient method for global optimisation that is able to balance between exploration and exploitation: when to prioritise between learning about the unknown and when to improve on the best known solution. We have applied this approach in a range of areas from optimisation of aerodynamic wings to designing small molecule drugs for biomedical applications. 
​Building and tuning deep neural networks for specific applications such as object detection and labelling. Using our IBM Power systems with NVIDIA GPUs, we can train complex neural networks much more rapidly than conventional compute resources.

Our Bayesian Optimisation tools are accessible through an API - Bayesian Optimisation as a Service (BOaaS), allowing end users to build their own applications without having to deploy the tools on premises.
Our team are able to develop image processing models for specific applications through collaborative R&D projects.

​​Applications

General purpose optimisation problems including: 
  • Product design
  • Process optimisation
  • Drug design
  • Deep neural network training​

​​Benefits

  • ​Increased efficiency of product operation and production
  • Reduced training time for image classification tasks


To learn how BayesOpt can be used to speed up search and optimisation problems – contact us

Contact: