Inventor · disambiguated record
Krishnakumar Narayanan Nair
Also filed as: NAIR KRISHNAKUMAR · NAIR KRISHNAKUMAR NARAYANAN
32 granted patents·12 pending applications·119 citations·filing 2016–2025
95Inventor score
Top patents by PatentIndex Score
44 records- 0195US11455143B2Using a low-bit-width dot product engine to sum high-bit-width numbersMETA PLATFORMS INC·Filed 2020·Granted Sep 27, 2022·6 cites·20 claims
- 0295US11138292B1Circuit and method for computing depthwise convolutionFACEBOOK INC·Filed 2019·Granted Oct 5, 2021·35 cites·20 claims
- 0394US11520853B2Mapping convolution to a partition channel convolution engineMETA PLATFORMS INC·Filed 2020·Granted Dec 6, 2022·5 cites·20 claims
- 0492US11599181B1Systems and methods for reducing power consumption of convolution operations of artificial neural networksMETA PLATFORMS INC·Filed 2019·Granted Mar 7, 2023·10 cites·20 claims
- 0591US10872038B1Memory organization for matrix processingFACEBOOK INC·Filed 2019·Granted Dec 22, 2020·12 cites·20 claims
- 0690US11537865B2Mapping convolution to a channel convolution engineMETA PLATFORMS INC·Filed 2020·Granted Dec 27, 2022·3 cites·20 claims
- 0790US11443013B2Pipelined pointwise convolution using per-channel convolution operationsMETA PLATFORMS INC·Filed 2020·Granted Sep 13, 2022·3 cites·20 claims
- 0889US11106430B1Circuit and method for calculating non-linear functions of floating-point numbersFACEBOOK INC·Filed 2019·Granted Aug 31, 2021·10 cites·20 claims
- 0986US11120328B1Systems and methods for reducing power consumption of convolution operations for artificial neural networksFACEBOOK INC·Filed 2019·Granted Sep 14, 2021·5 cites·20 claims
- 1082US12254061B2Apparatuses and methods to accelerate matrix multiplicationINTEL CORP·Filed 2018·Granted Mar 18, 2025·3 cites·25 claims
- 1182US11501147B1Systems and methods for handling padding regions in convolution operationsMETA PLATFORMS INC·Filed 2020·Granted Nov 15, 2022·2 cites·20 claims
- 1281US10079721B2Integrated digital network management platformNetsights360·Filed 2016·Granted Sep 18, 2018·14 cites·20 claims
- 1379US11054998B1High bandwidth memory system with distributed request broadcasting mastersFACEBOOK INC·Filed 2019·Granted Jul 6, 2021·3 cites·2 claims
- 1478US11481471B2Mapping convolution to a matrix processor unitMETA PLATFORMS INC·Filed 2019·Granted Oct 25, 2022·3 cites·20 claims
- 1576US11580192B2Grouped convolution using point-to-point connected channel convolution enginesMETA PLATFORMS INC·Filed 2020·Granted Feb 14, 2023·1 cites·20 claims
- 1676US11275560B2Hardware for floating-point arithmetic in multiple formatsMETA PLATFORMS INC·Filed 2020·Granted Mar 15, 2022·1 cites·20 claims
- 1775US10761757B2Apparatus and method for coherent, accelerated conversion between data representationsINTEL CORP·Filed 2018·Granted Sep 1, 2020·2 cites·24 claims
- 1873US11188303B2Floating point multiply hardware using decomposed component numbersFACEBOOK INC·Filed 2019·Granted Nov 30, 2021·1 cites·20 claims
- 1972US2024095304A1Flexible matrix processingMETA PLATFORMS INC·Filed 2023·Application pending·0 cites
- 2072US2023252263A1Point to point connected processing elements with data joiner componentsMETA PLATFORMS INC·Filed 2023·Application pending·0 cites
- 2171US11829441B2Device and method for flexibly summing matrix valuesMETA PLATFORMS INC·Filed 2022·Granted Nov 28, 2023·0 cites·20 claims
- 2270US11763131B1Systems and methods for reducing power consumption of convolution operations for artificial neural networksMETA PLATFORMS INC·Filed 2021·Granted Sep 19, 2023·0 cites·20 claims
- 2370US2023251903A1High bandwidth memory system with dynamically programmable distribution schemeMETA PLATFORMS INC·Filed 2023·Application pending·0 cites
- 2470US2024028905A1Artificial neural network training using flexible floating point tensorsINTEL CORP·Filed 2023·Application pending·0 cites
- 2567US11537301B2High bandwidth memory system with distributed request broadcasting mastersMETA PLATFORMS INC·Filed 2021·Granted Dec 27, 2022·0 cites·20 claims
- 2667US2025232002A1Apparatuses and methods to accelerate matrix multiplicationINTEL CORP·Filed 2025·Application pending·0 cites
- 2766US11379557B2Device and method for flexibly summing matrix valuesMETA PLATFORMS INC·Filed 2020·Granted Jul 5, 2022·0 cites·20 claims
- 2864US2022365784A1Matrix processing instruction with optional up/down sampling of matrixMETA PLATFORMS INC·Filed 2022·Application pending·0 cites
- 2964US2023004624A1High throughput matrix processor with support for concurrently processing multiple matricesMETA PLATFORMS INC·Filed 2022·Application pending·0 cites
- 3063US2023056304A1Using a low-bit-width dot product engine to sum high-bit-width numbersMETA PLATFORMS INC·Filed 2022·Application pending·0 cites
- 3162US11657252B2Point to point connected processing elements with data joiner componentsMETA PLATFORMS INC·Filed 2019·Granted May 23, 2023·0 cites·20 claims
- 3262US2022107782A1Floating point multiply hardware using decomposed component numbersFACEBOOK INC·Filed 2021·Application pending·0 cites
- 3361US11372644B2Matrix processing instruction with optional up/down sampling of matrixMETA PLATFORMS INC·Filed 2019·Granted Jun 28, 2022·0 cites·20 claims
- 3458US11663043B2High bandwidth memory system with dynamically programmable distribution schemeMETA PLATFORMS INC·Filed 2019·Granted May 30, 2023·0 cites·20 claims
- 3558US11409838B2High throughput matrix processor with support for concurrently processing multiple matricesMETA PLATFORMS INC·Filed 2019·Granted Aug 9, 2022·0 cites·20 claims
- 3658US2025190221A1Systems and methods for parallelization of embedding operationsADVANCED MICRO DEVICES INC·Filed 2024·Application pending·0 cites
- 3756US12205035B2Artificial neural network training using flexible floating point tensorsINTEL CORP·Filed 2018·Granted Jan 21, 2025·0 cites·25 claims
- 3853US11699081B2Systems and methods for reducing data movement during convolution operations in artificial neural networksMETA PLATFORMS INC·Filed 2019·Granted Jul 11, 2023·0 cites·20 claims
- 3953US11688032B2Three-dimensional convolution pipeline with memory organizer unitMETA PLATFORMS INC·Filed 2019·Granted Jun 27, 2023·0 cites·20 claims
- 4051US11520854B2Support for different matrix multiplications by selecting adder tree intermediate resultsMETA PLATFORMS INC·Filed 2019·Granted Dec 6, 2022·0 cites·20 claims
- 4148US11531619B2High bandwidth memory system with crossbar switch for dynamically programmable distribution schemeMETA PLATFORMS INC·Filed 2019·Granted Dec 20, 2022·0 cites·20 claims
- 4248US2021334072A1Mapping convolution to connected processing elements using distributed pipelined separable convolution operationsFACEBOOK INC·Filed 2020·Application pending·0 cites
- 4346US11640537B2Mechanism to perform non-linear functions in a machine learning acceleratorINTEL CORP·Filed 2019·Granted May 2, 2023·0 cites·16 claims
- 4444US2020364047A1High throughput neural network operations using inter-layer memory layout transformationFACEBOOK INC·Filed 2019·Application pending·0 cites
Identity basis: PatentsView inventor disambiguation (2025Q4-odp release). How scoring works →