Predicate |
Object |
assignee |
http://rdf.ncbi.nlm.nih.gov/pubchem/patentassignee/MD5_51d028c578ae85cb937b5b34a5129fbc http://rdf.ncbi.nlm.nih.gov/pubchem/patentassignee/MD5_8e0b0115c7643f86eafad4a975e6f41f http://rdf.ncbi.nlm.nih.gov/pubchem/patentassignee/MD5_b7844dc8eb295d8429f51b498c239775 http://rdf.ncbi.nlm.nih.gov/pubchem/patentassignee/MD5_064dffc5aa01f244c7ec643dc8386d2e http://rdf.ncbi.nlm.nih.gov/pubchem/patentassignee/MD5_8fed1c7a93af2cc7c0cbe259985fe6a5 |
classificationCPCAdditional |
http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-048 |
classificationCPCInventive |
http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06F17-16 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06F18-217 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N5-046 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-08 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-084 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-045 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-048 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-04 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N20-10 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N10-00 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06F18-214 |
classificationIPCInventive |
http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06N99-00 |
filingDate |
2018-05-31-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
inventor |
http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_c7fefc2cffa1329953754a3b46f60d84 http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_cc0da10632d604aaf29d3741b01b2f97 http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_ab627a26cb63f12eb6065d6e98eb884f http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_6f93de9d457d8e620fd474c732912a6f |
publicationDate |
2018-12-06-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationNumber |
WO-2018222896-A1 |
titleOfInvention |
Gradient-based training engine for quaternion-based machine-learning systems |
abstract |
A deep neural network (DNN) includes hidden layers arranged along a forward propagation path between an input layer and an output layer. The input layer accepts training data comprising quaternion values, outputs a quaternion-valued signal along the forward path to at least one of the hidden layers. At least some of the hidden layers include quaternion layers to execute consistent quaternion (QT) forward operations based on one or more variable parameters. A loss function engine produces a loss function representing an error between the DNN result and an expected result. QT backpropagation-based training operations include computing layer-wise QT partial derivatives, consistent with an orthogonal basis of quaternion space, of the loss function with respect to a QT conjugate of the one or more variable parameters and of respective inputs to the quaternion layers. |
isCitedBy |
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-111782219-A http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-110069985-A http://rdf.ncbi.nlm.nih.gov/pubchem/patent/US-2020117993-A1 http://rdf.ncbi.nlm.nih.gov/pubchem/patent/EP-3690756-A1 http://rdf.ncbi.nlm.nih.gov/pubchem/patent/US-11521060-B2 http://rdf.ncbi.nlm.nih.gov/pubchem/patent/US-11263526-B2 http://rdf.ncbi.nlm.nih.gov/pubchem/patent/US-11593643-B2 http://rdf.ncbi.nlm.nih.gov/pubchem/patent/US-10733511-B1 |
priorityDate |
2017-05-31-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
type |
http://data.epo.org/linked-data/def/patent/Publication |