Skip to main content

Monetization

Rizemind combines state-of-the-art federated learning with rigorous, game-theoretic reward allocation to ensure that every trainer’s contribution is fairly valued—and transparently rewarded using Rizenet—so as to drive broad participation and collaborative model improvement. Our Federated Learning framework enables models to learn from distributed datasets without compromising privacy or data security—and ultimately fostering widespread adoption and collaborative growth.

Shapley Value​

Understanding Shapley Value​

To incentivize ongoing participation, Rizemind employs the Shapley value from cooperative game theory, which gives a principled way to split the total “gain” of collaboration among contributors. The Shapley value, originating from cooperative game theory, provides a robust method for fair reward calculation and distribution among contributors based on their individual impact. By calculating the average marginal contribution of each participant across all possible collaboration scenarios, the Shapley value ensures fairness and transparency in reward allocation. In the FL context, the “value” of a coalition is the global model’s performance (e.g. accuracy) when trained only by that subset of trainers. Shapely value is the only payment rule that satisfies the four properties of Efficiency, Symmetry, Linearity, and Null player:

  • Efficiency: All of the overall improvement is distributed; the sum of individual Shapley values equals the full coalition’s gain.
  • Symmetry: Two trainers who contribute identically receive equal rewards, regardless of labeling or ordering.
  • Linearity: If you combine two separate training tasks, each trainer’s reward is the sum of their rewards for each task individually.
  • Null player: A trainer whose participation never changes model performance gets a Shapley value of zero.

In addition, it has the feature of Anonymity, in which the labeling of the agents doesn't play a role in the assignment of their gains.

Applications of Shapley Value​

Shapley value has wide-ranging applications, especially in contexts requiring fair resource allocation, such as profit-sharing, cost division, and contribution measurement in collaborative projects. It is particularly valuable in complex ecosystems like Federated Learning, where understanding the precise contribution of each trainer significantly impacts overall trust and participation.

Ensuring Fair Contribution Assessment​

Rizemind employs the Shapley value to quantify the marginal contributions of each trainer by systematically evaluating what is the best candidate for the global model. Utilizing Flower's Federated Evaluation, Rizemind effectively distributes the assessment tasks across trainers, ensuring rapid and accurate calculations. This parallelized evaluation process directly addresses data heterogeneity, ensuring fairness even in highly diverse training scenarios.

Trust and Reliability​

By utilizing the Rizenet blockchain, we provide a transparent environment for participants in the training procedure which utilizes precise measurements of the contributions to monetize Federated Learning. Our solution is engineered to maximize the potential of transparent nature of blockchain and the distributed and privacy of Federated Learning. We provide a robust integration of game theory methodologies ensures fairness, enhancing collaboration among trainers and encouraging continuous participation.

Last updated on