Federated learning requires more than Big Tech infrastructure

Making federated learning work across company boundaries requires more than a model capable of learning from diverse datasets. It requires a secure technology stack that can operate within each participant’s computing environment and coordinate training without exposing proprietary records.

That stack is becoming more sophisticated, reducing the technical risks of participation. Some biopharmas are already joining federated networks. How far participation will extend remains an open question, however, particularly when making strategically valuable data available for federated learning allows other participants to benefit from the resulting models...