US 11,940,958 B2
Artificial intelligence software marketplace
Michele M. Franceschini, White Plains, NY (US); Ashish Jagmohan, Irvington, NY (US); Kanthi Sarpatwar, Elmsford, NY (US); Karthikeyan Shanmugam, Elmsford, NY (US); and Roman Vaculin, Bronxville, NY (US)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Mar. 15, 2018, as Appl. No. 15/922,020.
Prior Publication US 2019/0287027 A1, Sep. 19, 2019
Int. Cl. G06F 16/18 (2019.01); G06F 16/13 (2019.01); G06F 16/182 (2019.01); H04L 9/06 (2006.01); H04L 9/00 (2022.01)
CPC G06F 16/1837 (2019.01) [G06F 16/137 (2019.01); G06F 16/1805 (2019.01); G06F 16/1865 (2019.01); H04L 9/0643 (2013.01); H04L 9/50 (2022.05)] 16 Claims
OG exemplary drawing
 
1. A computing system, comprising:
a network interface configured to communicate with nodes in a network; and
a processor configured to:
receive, via a blockchain ledger, a plurality of requests from a plurality of producer nodes, where the plurality of requests comprise a plurality of bids for training an iteratively-trained model available to the plurality of producer nodes;
select, from among the plurality of bids, a bid that comprises a hashed improvement value that is generated by a producer node hashing a respective improvement value using a locality-preserving hash function that hides the respective improvement value but preserves information for deriving the respective improvement value from the hashed improvement value;
execute an improvement value determination function on the hashed improvement value to locally determine an improvement value based on the information that is preserved by the locality-preserving hash function;
determine that the locally determined improvement value is equivalent to the hashed improvement value in the selected bid;
in response to the determination, retrieve hashed training data of the producer node associated with the selected bid; and
aggregate the hashed training data of the producer node with a summary of hashes of training data from previous iterations of the training model to generate an updated summary; and
store the updated summary via a data block of the blockchain ledger.