Patent application: Source Selection based on Diversity for Transfer Learning

• Select source model for transfer learning from multiple available source domains with little to no data in target domain. Automatic and highly scaleable due to only looking at diversity which is a marginal quantity.
• In addition to the patent application there is a paper published on this: H. Larsson, J. Taghia, F. Moradi and A. Johnsson, "Source Selection in Transfer Learning for Improved Service Performance Predictions," 2021 IFIP Networking Conference (IFIP Networking), 2021, pp. 1-9, doi: 10.23919/IFIPNetworking52078.2021.9472818.

Input(s):

• Candidate source ML models
• Candidate source data sets

Main feature(s):

• Automated selection of source model for transfer learning

Output(s):

• Selected source ML model

None, this is an intellectual property.

People/software responsible for ML model management.

Commercial license to be negotiated.