Fig. 5From: Neural network training with limited precision and asymmetric exponentNeural network accuracy in various exponent and mantissa bit count scenariosBack to article page