Logarithm-approximate floating-point multiplier is applicable to power-efficient neural network training
TaiYu Cheng, Yukata Masuda, Jun Chen, Jaehoon Yu, Masanori Hashimoto
Topics & Concepts
Computer scienceServerLogarithmArtificial neural networkMultiplier (economics)Floating pointField-programmable gate arrayElectrical efficiencyEdge deviceComputationComputer engineeringComputer hardwareParallel computingAlgorithmPower (physics)Cloud computingArtificial intelligenceMathematicsComputer networkOperating systemMathematical analysisMacroeconomicsQuantum mechanicsEconomicsPhysicsAdvanced Neural Network ApplicationsStochastic Gradient Optimization TechniquesFerroelectric and Negative Capacitance Devices