search

iamtrask / Sonar

DEPRECATED - Decentralized Machine Learning Server (hosted on Blockchain)
Apache License 2.0
141 stars 49 forks source link

Corrected Incentive Calculation #38

Closed bendangnuksung closed 6 years ago

bendangnuksung commented 6 years ago

current incentive calculation is not accurate because it is calculating the incentive without considering 'Model.target_error'. Its assumes the target_error =0. Here it considers the 'target_error' to calculate incentive.

anoff commented 6 years ago

Not sure if anyone actually put any thoughts into how the bounty should really be calculated.

Please make sure the PR does not break the build.

bendangnuksung commented 6 years ago

if a new_model reduces error lesser than the target_error, what will its incentive be?

bendangnuksung commented 6 years ago

Closing as there is conflict in ModelRepository.sol, will issue a new PR.