Data Obfuscation Scenarios for Batch ELM in Federated Learning Applications
摘要
The batch formulation of the Extreme Learning Machines (ELM) method fits well with federated learning scenarios. This paper proposes and investigates the strategies for data obfuscation that can be used in combination with ELM to create a secure distributed learning environment. Results show that the model allows for significant levels of added noise with minimal impact on its predictive performance; enabling secure federated learning in tasks that can benefit from it.