Development of an innovative tool for automated and optimized selection of computing resources for machine learning experimentation
The project concerns the development of innovative methods and tools, expanding the Deepstributed Platform’s operation by optimized management of computing resources. Optimal dimensioning of resources allocated for Platform’s tasks will be enabled by predicting the value of performance index for available computational resources, based on the characteristics of the resourcesand performed machine learning experiments.
The project includes R&D works in the field of:
– developing new computing resource models, their allocation and queuing methods;
– developing machine learning prediction model of performance index of computational resources, methodologies and procedures associated with this model (e.g., continual learning);
– developing and evaluating scenarios for using the platform in the new configuration;
– integration of individual elements with the platform;
– testing in near real-life conditions and optimizing the platform architecture.
A comprehensive solution will be developed to reduce costs of machine learning experiments – from the configuration, commissioning and management of experiments, to management and dimensioning of computing infrastructure – optimized resource allocation for performed machine learning experiments, taking into account its nature and functional/performance characteristics of the corresponding computing resource. The main recipients of the product will be machine learning engineers (MLE), for which the developed tool will be support in their everyday work. Because of the unique platform features, the basic customer group will be entities operating on the machine learning market employing MLE (AI Software Houses, research and development units of industrial entities, universities), which in addition to the possibilities optimizing the costs of conducting experiments, will be able to share their resources within the Deepstributed Platform.
Project value: 3 200 017,06 PLN
Contribution of European Funds: 2 478 873,35 PLN