What is federated machine learning? A 101 guide

Why is it needed?

To secure a competitive edge, companies today need access to diverse, extensive, high-quality data that enhances the performance of their AI systems compared with that of their rivals. But concerns over privacy can limit the use of unique, relevant data for analysis.

This problem can be alleviated by means of privacy-preserving federated learning. In combination with a special type of encryption, this enables an AI model (or any other type of algorithm) to be trained using data from multiple, decentralized servers controlled by different organizations while respecting the privacy of the individuals or organizations whose data is being used for the training.

Simply put, federated learning entails sending the algorithm to the data, rather than sending the data to the algorithm.

How does it work across and within industries?

The value of federated learning collaborations stems from the ability to train AI systems on much richer datasets than any one organization could assemble alone. To do this, you need to identify partners whose data could be used in a federated learning approach.

While it might appear to be more logical for organizations from different industries to collaborate than companies in the same industry, federated learning can facilitate cooperation within industries – including between direct competitors.

How do we assess which external organizations to partner with?

You need to consider the nature of the data that your organization brings to the collaboration. The data used to train a given model must contain a large number of samples (clients, patients, insurance policies, or whatever the system aims to understand) and a large number of features (variables) within each sample.

Six main factors to consider when deciding whom to collaborate with

1. The state of your data

Before searching for external partners, you must determine whether your data is poor, vertical, horizontal, or rich (see Know Your Data to Harness Federated Machine Learning).

2. The structure of their data

Organizations with vertical data should look outside their industry. Those with horizontal data should seek collaborations within their own industry – and even with competitors.

3. A logical starting point

Start with one algorithm and a trusted partner (this could be within the organization) to explore federated learning securely.

4. The potential for data monetization

Federated learning is a privacy-centric method to monetize your data by contributing to the AI training processes of others. Identifying partner organizations with a need for this data is crucial to such monetization efforts.

5. Possible technical challenges

Anticipate technical challenges and costs, such as harmonizing data formats and structures between different organizations.

6. Employee buy-in

Work with people, not around them. Federated learning needs a transformational approach. Get widespread buy-in and clearly assign roles and resources.