2PM.Network
WebsiteXDiscordLinkedinGithub
  • Overview
    • What is 2PM.Network
    • Architecture
    • FAQ
    • Official Social Channels
  • 2PM Data VSIES Service
    • What is Data VSIES and why is it important
    • [V] Data Validation (ZK)
    • [SI] Data Standardization and Index
    • [E] Data Encryption Client (FHE)
    • [S] Data Storage and Access
    • Data VSIES SDK
  • Node Framework
    • Modular Architecture
    • Federated Learning
      • Horizontal Federated Learning Task
      • Logistic Regression Task
      • On-chain Secure Aggregation
      • Typical Scenarios
    • FHE Machine Learning
      • Built-in Models
      • Deep Learning
      • Typical Scenarios
    • Task Submission
    • Running a 2PM Node
      • Installation
      • Chain Connector Configuration
      • Data Preparation
      • Joining a Subnet
  • Security and Verification
    • Node Staking and Slash Mechanism
    • Running Verification Client
      • EigenLayer
      • Mind Network
    • Restaking and Delegation
  • Model Inference
    • 2PM Node Inference API
    • Posting Request to a Subnet Model
    • Getting Inference Results on Chain
      • Oracle Adapters
  • Monetization and Incentives
    • AI Model IP Assets
    • Distribution Algorithm
  • 2PM DAO
    • Build Subnets
      • Establishing New Subnets
      • General Requirements
      • Data Schema Definition
      • Model Selection
      • Task Implementation
    • $DMP Token
  • Deployed Smart Contracts
    • Subnets on Testnets
    • Local Deployment Guideline
  • Ecosystem
    • Partners
    • Use Cases
      • Private Personalized Recommendation
Powered by GitBook
On this page
  • Distribution Algorithm and Process
  • Computing Power Contribution
  • Data Contribution
  • Royalty Vault Mechanism
  1. Monetization and Incentives

Distribution Algorithm

Distribution Algorithm and Process

Distribution Process: Royalty Tokens are distributed by the Royalty Vault, which retrieves each user or node’s data contributions from either the FHE Task Contract or the Federated Learning Task Contract. The contributions of computing power and data are weighted in a 7:3 ratio, respectively.

Computing Power Contribution

Contribution Process and Algorithm: In the computation rounds, the token distribution follows an exponential decay model to ensure continuous participation from nodes, even in later rounds. The model encourages newer models to attract fresh participation, sustaining the ecosystem’s vibrancy.

The exponential distribution used for computing contributions is defined as:

f(x∣λ)=λe−λx,x≥0f(x|\lambda) = \lambda e^{-\lambda x}, \quad x \geq 0f(x∣λ)=λe−λx,x≥0

Here, for computing power is significantly smaller than for data contributions, acknowledging the consistent utility of computing power throughout the lifecycle of the subnet.

Data Contribution

Contribution Process and Algorithm: In scenarios involving Fully Homomorphic Encryption, the computational contribution is straightforward since only a single node participates per round. This node receives the full share of computational rewards. Data contributors, on the other hand, have their contributions verified during the encryption process by contracts, which records the contributor’s address in the 0G storage node.

Data contributions are incorporated into the Royalty Token distribution as follows:

DKL(P∥Q)=∑iP(i)log⁡P(i)Q(i)D_{KL}(P \parallel Q) = \sum_{i} P(i) \log \frac{P(i)}{Q(i)}DKL​(P∥Q)=i∑​P(i)logQ(i)P(i)​

for discrete distributions, and

DKL(P∥Q)=∫−∞∞p(x)log⁡p(x)q(x),dxD_{KL}(P \parallel Q) = \int_{-\infty}^{\infty} p(x) \log \frac{p(x)}{q(x)} , dxDKL​(P∥Q)=∫−∞∞​p(x)logq(x)p(x)​,dx

for continuous distributions. The Kullback-Leibler divergence measures how one dataset diverges from another, establishing a basis for the contribution-based rewards.

Royalty Vault Mechanism

The FHE Task Contract provides the Royalty Vault with necessary proofs (including data digest, user addresses, original data upload transaction details like hashes, and proofs of computation for the data). These proofs detail the addresses and contributions noted in each task contract’s record. The Royalty Vault then calculates the tokens each user is entitled to claim based on these contributions and whether they have been claimed before.

This model ensures that contributors of both data and computing power are fairly compensated for their participation, promoting a balanced growth and sustainability within the 2PM Network's ecosystem.

PreviousAI Model IP AssetsNextBuild Subnets

Last updated 10 months ago