http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-109359735-A
Outgoing Links
Predicate | Object |
---|---|
assignee | http://rdf.ncbi.nlm.nih.gov/pubchem/patentassignee/MD5_d6a6f422b091ba12ea61d4adbf1b0e8e |
classificationCPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-045 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-063 |
classificationIPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06N3-063 http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06N3-04 |
filingDate | 2018-11-23-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
inventor | http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_094e8c73d5435f637dd3918641c01511 http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_62419822b88810329a46878395e3fa94 http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_d3a213df2796885cbecd4f8f9ab8f64e |
publicationDate | 2019-02-19-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationNumber | CN-109359735-A |
titleOfInvention | Data input device and method for hardware acceleration of deep neural network |
abstract | The invention provides a data input device accelerated by deep neural network hardware, comprising a PE array group, triple input buffer, weight buffer, output buffer and controller; the triple input buffer is connected with the PE array group in one-to-one correspondence, and the PE array The groups and output buffers are connected in one-to-one correspondence; each triple input buffer is connected to the input feature map storage area through a bus; the controller is connected to the input feature map storage area and each triple input buffer respectively; the weight buffer is connected to the weight storage area through a bus and the weight buffer is connected to each PE array group; each output buffer is connected to the output feature map storage area through a bus. The present invention also provides a data input method using the above-mentioned device for hardware acceleration of deep neural network. Through the design of triple input buffer, the purpose of hiding data transmission time delay is realized, and the overlapping part of data can be directly reused, thereby improving the efficiency of data transmission. Data reuse rate. |
isCitedBy | http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-110705687-A http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-110390383-B http://rdf.ncbi.nlm.nih.gov/pubchem/patent/WO-2020258841-A1 http://rdf.ncbi.nlm.nih.gov/pubchem/patent/WO-2020258527-A1 http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-110390383-A |
priorityDate | 2018-11-23-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
type | http://data.epo.org/linked-data/def/patent/Publication |
Incoming Links
Total number of triples: 23.