Decentralized Machine Learning

Decentralized Machine Learning Revolutionize the Use of Untapped Data with Privacy Protection

Decentralized Machine Learning (DML) is developing a decentralized protocol with the aid of blockchain, machine learning and federated learning technology etc. to unleash the potential of privately owned data that are currently not being fully utilized for machine learning and to drive innovation from periphery by forming a developer community in the marketplace.

What is Decentralized Machine Learning (DML)?

DML aims to develop a decentralized protocol that connects potentially billions of smart devices (e.g. smartphones, tablets etc.) for machine learning algorithms to be run on the untapped privately owned data that are stored in each device. The protocol will run machine learning algorithms directly on the device without any data extraction to ensure data privacy protection and also better utilize the processing power of each device.

Not only does DML better utilize the untapped data and underutilized processing power of each device to advance machine learning, but it also aims to drive innovation from periphery for machine learning development. An algorithm marketplace will be created to facilitate talented developers from all over the globe to publish their algorithms, hence machine learning development will not be contributed by a confined group of core developers within large corporations.

How DML changes the Machine Learning Ecosystem?

In this article, we focus discussing the decentralized machine learning from a data perspective, which is one of the critical elements for the success of machine learning development.

Avoid Oligopoly to control Majority of Available Data

Fascinating by the vivid growth in machine learning development, companies have been applying machine learning to their business such as anticipating customer preference and improving media purchase.  Tech giants such as Facebook, Amazon, Google etc. have heavily invested in machine learning research and data acquisition, they gain control and profit from the majority of data that users are unintentionally and passively provided without compensation. Data owners should have the rights to control their own data and be rewarded for contributing their data for machine learning development.  DML protocol allows data owners to have control over their own data by authorizing the machine learning algorithms to be run on specific types of data according to their will and receive compensations according to their contribution.

Unleash Massive Market Potential by Utilizing Untapped Data

Furthermore, the existing data that are acquired by these tech giant and being used for machine learning are just tip of the iceberg. Just count how many photos in your albums have you uploaded to the social media platforms? How many purchasing website or applications will you use to buy different types of products? Although the leading tech giants gain access and control of the majority of available data, there are still a huge amount of untapped data, which is located at the bottom of the iceberg. With the aid of DML protocol, the untapped private data can be better utilized to advance machine learning development.

Privacy Protection through on-device Machine Learning

DML enables on-device machine learning, hence algorithms will be run directly on the devices without the need to extract and store the raw data in cloud or third-party platform. Data privacy can be well protected.

Win-win Ecosystem

Data owners can contribute to machine learning effort and monetize their valuable data with full control and privacy protected, which are currently unfeasible. For corporations or individuals, who wish to apply machine learning to their business, they are given an alternative source of machine learning models with better quality of untapped data to be applied.

Far Beyond Untapped Data Application

DML is far more than a decentralized data project, it also creates a platform to promote innovation from periphery for machine learning evolvement. As a result, machine learning advancement will not be confined to the contributions and creativity of a small group of core developers, who worked in large corporation and bounded by bureaucratic guidelines.

More about DML

To study deeper about DML protocol, you can refer to their website:, whitepaper, introductory video and other materials.

This is a sponsored press release and does not necessarily reflect the opinions or views held by any employees of The Merkle. This is not investment, trading, or gambling advice. Always conduct your own independent research.