Success Story - en

Preliminary Study on NN Comparison

The project team had to conduct a full review of the scientific literature on methods for effectively training neural network models, assessing their quality, ranking models according to their quality within the framework of the NAS task, which can reduce the number of used computing resources by several times.

Methods for ranking models according to their quality significantly speed up the selection of architectures during automatic search, reducing the complete search procedure to several GPU hours. Methods for effective training and assessing their quality allow you to quickly obtain the final quality of the model, which leads to faster development of the final model and allows you to reduce the amount of necessary computing resources to solve the problem.

Team decision:
The laboratory team analyzed more than 70 works, based on 13 of which they formulated a vision for the project and formed a work plan for the implementation of the required models.

  • An overview of the area has been compiled
  • Directions for model development have been selected
Research Compression
Made on