Machine learning has made huge strides in recent years, allowing organizations to train models that can recognize patterns in vast amounts of data that, until now, have been difficult to navigate.
But, as the amount of data grows, so does the challenge of processing it; federated learning is a decentralized approach to machine learning that aims to help solve this problem.
For federated learning to be adopted, data custodians must make data available without sharing that data or moving it to a central location.
In doing so, they can allow data scientists and machine learning engineers to run advanced analytics and train models in a federated (ie. distributed) manner.
Thanks to technological advances, large corporations and startups alike are working to make federated learning ubiquitous so that companies – and consumers too – can benefit.
According to a new research report from Polaris Market Research, the global federated learning market was valued at USD 110.82 million in 2021 and is expected to reach USD 266.77 million by 2030, growing at a CAGR of 10.7% throughout the forecast period.
A Powerful Present
Unlike machine learning techniques that utilize centralized data, stored in a silo, federated learning operates differently by enabling collaborative learning among distributed data sources without pooling data centrally.
Although still in the early stages, federated learning has already experienced remarkable growth and innovation, signaling its arrival as a viable solution with a highly attractive value proposition.
This includes its ability to help organizations comply with ever-increasing data privacy and AI regulation, which typically dictates that sensitive data cannot be moved from the territory it was originally created in.
It achieves this by allowing computational access to data while custodians remain in full physical and operational control of their data, allowing them to comply with data privacy regulation requirements and protect their intellectual property (IP) without requiring them to share or centralize it.
By removing the need to move data outside of their environment, security is greatly improved while the risk of data breaches, IP leaks, and unauthorized data access is reduced.
Not only is this beneficial from a privacy standpoint, it offers many additional benefits; the cost of copying and centralizing data is completely removed, while accuracy is drastically improved thanks to the availability of previously inaccessible data that improves the accuracy of ML models or analytics.
Combined, this has led to rapid innovation and growth in a plethora of industries. Medical imagery, for example, is a data-intensive source that benefits from federated-learning-powered AI.
By harnessing the power of AI, medical professionals can achieve a new level of analytical depth and improve the efficiency and precision of image processing.
Radiologists, in particular, face an exponential increase in their workload, and often need to read medical images in a matter of a few seconds to meet the demands on their time.
This intensity can negatively affect their mental state and concentration, leading to reduced diagnostic accuracy, additional stress for patients, and burnout for healthcare specialists.
Leveraging federated learning to address these challenges, AI-driven applications can be developed that are capable of improving diagnostic accuracy while providing workflow support previously thought impossible.
Moreover, AI can detect abnormalities in medical images that the human eye might miss or misinterpret, unlocking new possibilities for early detection and treatment.
Medtech, in general, is also benefiting from federated learning.
For example, by training and evaluating language models on sensitive medical information and health data, telemedicine and patient apps are developed without exposing personal data and are more accurate, predictable and trustworthy as a result of access to more complete, representative data.
In healthcare, AI has the potential to revolutionize the way medical professionals treat patients. For instance, it can be used to create models that accelerate the drug discovery process, the development of diagnostics, or clinical decision support applications, ultimately leading to better diagnosis and treatments.
However, the most critical opportunity lies in leveraging data collaboration to accelerate the industry's journey towards helping more patients. Traditional approaches, for instance in drug discovery, require such a high amount of financial resources, that only a limited amount of diseases can be targeted while staying economically viable.
Data collaborations can lead to much better models providing more reliable and precise predictions and lower costs - allowing pharma corporations to widen their strategy from billion dollar therapeutics to smaller markets and helping a wider range of patients.
Now, let’s take a look at the top 5 federated learning startups.
Top 5 federated learning startups
Apheris
Apheris’ product, the Compute Gateway, connects distributed data via a federated computing infrastructure and adds additional governance, security and privacy controls, ensuring that only approved computations can be executed on the data.
During the whole process, data owners stay in physical and operational control of their data, while data and model IP stay protected and compliance requirements (such as for GDPR) are fulfilled.
Key benefits of Apheris' Compute Gateway include:
Rapid market adaptation: It significantly shortens the time from data access to actionable insights, streamlining compliance and operational scalability. This agility accelerates revenue generation and opens new growth opportunities through innovative use cases.
Enhanced data usage: The Compute Gateway provides direct computational access to diverse, multi-modal datasets while safeguarding sensitive information. This capability maximizes data value and supports advanced analytics.
Minimized operational risk: The product integrates a flexible governance framework that enables Data Custodians to stay in compliance with global regulations, such as GDPR. Strong data security and IP protection measures ensure that data remains secure and private, mitigating risks of breaches and unauthorized access.
Increased breadth and depth of data: The Compute Gateway supports a wide range of data modalities and facilitates collaboration across multiple sites and regions. This enhanced coverage helps organizations differentiate themselves and create data networks.
Faster insights and advanced analytics capabilities: By providing a programmatic interface and integrating with AI technologies, the Compute Gateway enables organizations to generate outcomes and use sophisticated analytics tools. This capability enhances research efficiency and meets the growing demand for complex data analysis.
NVIDIA FLARE
NVIDIA FLARE (FLARE) is maintained by the one and only NVIDIA team. The federation engine has a tremendous amount of well designed (security & governance relevant) features, a security-hardened architecture and is domain-agnostic.
Furthermore, FLARE makes it easy to use models from MONAI and Hugging Face and enables ML Engineers to easily connect to existing ML workflows (PyTorch, RAPIDS, Nemo, TensorFlow).
Flower Labs
Flower Labs provides an open-source framework for federated learning, designed to support AI model training across decentralized data sources.
Flower is versatile, compatible with major machine learning frameworks such as PyTorch, TensorFlow, and Hugging Face, and enables scalability across cloud, mobile, and edge environments.
While Flower allows companies like Banking Circle, Nokia, Porsche, and Brave to enhance AI models without moving sensitive data, it lacks federated governance mechanisms for managing complex data privacy and compliance rules.
Federated Learning and its Bright Future
Federated learning has the potential to revolutionize the way we think about machine learning by allowing organizations to train models with greater accuracy and privacy. By decentralizing the process, federated learning also gives companies greater control over their data.
It’s evident that the time has come for organizations to embrace federated learning as a tool that’s able to unlock rapid innovation and significant growth.
What is federated ML and analytics and why does it matter?