Other Technical Features
AIRTIST can be enhanced in the future to incorporate automatic machine learning (AutoML) capabilities. AutoML aims to automate the design and training of DNNs and other AI models. The basic idea is to use a certain encoding to capture the overall solution space of DNNs and then search for a solution in the space to best meet a given objective.
As an exhaustive traversal of the solution space is infeasible, adopting an efficient search algorithm like reinforcement learning or evolutionary algorithm is essential. AutoML can generate high-quality models at the cost of intensive computation.
With AutoML, AIRTIST can be constructed through automatic design and optimisation by leveraging the distributed computing network. Such an approach revolutionises the process of building MASTER/MAESTRO by allowing users without AI backgrounds to participate in the ecosystem.
Copy link