Use cases
Proven applications with the following type of problems.
- estimation of values previously only accessible by expensive, empirical, experimentation
- ability to estimate the endpoints in complex, multistage, multi-ingredient processes
- qualification of estimates by robust and meaningful quality metrics indicative of uncertainty
- ability to identify and correct outlier data and to suggest empirical experiments that will improve overall uncertainty of the model
- computationally efficient and scalable from small matrices to big data
- large amounts of incomplete anonymised, numerical data
- numerical data combined with models or graph functions
Implementation and delivery
Alchemite™, can be used standalone to generate models or easily be integrated into existing software stacks and workflows.
Alchemite enables the configuration of a number of hyper-parameters to specify, for example, number of hidden layers, time to train or level of accuracy recquired.
The tool is easily configured through a simple parameter file and data is managed through simple data files
A management console can be delivered as a fully managed, on demand solution or deployed internally on a per user or enterprise license basis.
The interface allows for non-technical users to easily train and deploy networks, using local or cloud based compute resources to manage jobs. Trained models can then be used through a number of easy to use interfaces or shared and tested via public facing API's.