Deep Neural Network Optimized to Resistive Memory with Nonlinear Current-Voltage Characteristics
This addresses a hardware-specific bottleneck for improving energy-efficient AI computation using emerging nonvolatile memory, representing an incremental advance in optimizing neural networks for resistive memory.
The paper tackles the problem of accuracy loss in neural networks due to nonlinear current-voltage characteristics in resistive memory crossbar arrays by proposing a method to reconstruct neural networks optimized for these arrays, resulting in significantly higher inference accuracies on MNIST and CIFAR-10 datasets compared to conventional networks.
Artificial Neural Network computation relies on intensive vector-matrix multiplications. Recently, the emerging nonvolatile memory (NVM) crossbar array showed a feasibility of implementing such operations with high energy efficiency, thus there are many works on efficiently utilizing emerging NVM crossbar array as analog vector-matrix multiplier. However, its nonlinear I-V characteristics restrain critical design parameters, such as the read voltage and weight range, resulting in substantial accuracy loss. In this paper, instead of optimizing hardware parameters to a given neural network, we propose a methodology of reconstructing a neural network itself optimized to resistive memory crossbar arrays. To verify the validity of the proposed method, we simulated various neural network with MNIST and CIFAR-10 dataset using two different specific Resistive Random Access Memory (RRAM) model. Simulation results show that our proposed neural network produces significantly higher inference accuracies than conventional neural network when the synapse devices have nonlinear I-V characteristics.