US 12,468,733 B2
Machine learning for optimized learning of human-understandable logical rules from medical or other data
Francesco Alesiani, Heidelberg (DE); and Markus Zopf, Heidelberg (DE)
Assigned to NEC CORPORATION, Tokyo (JP)
Filed by NEC Corporation, Tokyo (JP)
Filed on Sep. 7, 2023, as Appl. No. 18/462,454.
Application 18/462,454 is a continuation of application No. 17/668,443, filed on Feb. 10, 2022, granted, now 11,822,577.
Claims priority of provisional application 63/248,611, filed on Sep. 27, 2021.
Prior Publication US 2023/0418840 A1, Dec. 28, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 16/28 (2019.01); G06N 3/04 (2023.01)
CPC G06F 16/284 (2019.01) [G06N 3/04 (2013.01)] 16 Claims
OG exemplary drawing
 
7. A computing device configured for learning and applying a rule set from relational data, the device comprising:
one or more processors; and
a memory storing instructions, wherein the instructions when executed by the one or more processors cause the network device to implement a machine learning method of learning and applying a rule set from relational data, the method comprising:
receiving a graph representing relational data and a partial graph representing a part of the graph, wherein nodes represent elements of the graph, and edges represent relationships between nodes;
learning optimized logical rules that define the nodes and edges of the graph by:
defining a maximum satisfiability (MAX-SAT) problem for the graph;
estimating a gradient around a solution of the MAX-SAT problem for the graph to generate an intermediate representation of the graph by mapping features of the nodes and edges of the graph to an intermediate vector representation, and to produce the optimized logical rules, wherein the intermediate vector representation includes binary values and/or probabilistic values; and
inputting the graph to a static encoder, inputting the partial graph to a learned encoder, and propagating the estimated gradient to the static encoder from the learned encoder;
applying the optimized logical rules to a new graph; and
checking a validity of the new graph for satisfying the logical rules.