FTL sits at the intersection of federated learning — which coordinates distributed model training without data sharing — and transfer learning, which enables knowledge reuse across domains with differing feature spaces or data distributions. Standard federated learning assumes participating nodes share overlapping features; FTL specifically addresses scenarios where nodes do not have many overlapping features and users, as established in a foundational 2021 survey in this dataset.

Three dominant technical sub-domains are evident in the retrieved records: cross-silo federated aggregation (coordinating model updates between organizational participants using global aggregators), federated multi-task and personalized learning (training distinct but related models for different operational tasks), and knowledge transfer and domain adaptation (moving learned representations between federated endpoints with non-overlapping feature spaces).

In this dataset, IBM holds 6 patent documents (US, WO, DE) covering FL orchestration, trusted aggregation, and framework building. Huawei Technologies holds 4 patent documents (WO, CN, US, EP) spanning DLT-enabled aggregation and Bayesian distribution transfer. Telefonaktiebolaget LM Ericsson holds 4 patent documents focusing on distributed ML/FL node orchestration and vertical federated learning in 5G networks. Nokia Technologies Oy holds 3 patent documents for federated multi-task learning in network automation.

Koninklijke KPN N.V.’s 2026 WO patent introduces a parameter mapping mechanism to link model parameters between source and target device sets, including splitting and merging operations to handle structural heterogeneity. This is directly applicable to factories running different PLC and sensor configurations where model architectures differ between sites.

For multi-factory deployments where no single factory should act as the trusted aggregator, IBM’s 2024 DE patent uses Trusted Execution Environments (TEEs) and encrypted virtual machines to decentralize aggregation, allowing each factory to verify the aggregator’s authorization before contributing model updates. Huawei’s 2023 WO patent integrates Distributed Ledger Technology (DLT) to select aggregator entities and upload global FL model update data, enabling trustless and auditable cross-factory model governance.

The dataset reveals strong coverage of FTL mechanisms in telecom, healthcare, and general-purpose ML, but limited patent activity explicitly claiming multi-factory manufacturing as the primary use case. R&D teams in industrial AI have an opportunity to file domain-specific claims combining FTL with manufacturing process ontologies, equipment heterogeneity handling, and production KPI objectives, according to the strategic analysis in this dataset.