Fig. 1. (Color online) Schematics of flash-based CIM architecture. The pulse time of V_g and the threshold voltage is individually mapped as vector and matrix, then the amount of charge can represent the result of MVM.

Fig. 2. (Color online) (a) Schematic of adopted CHEI and HHI programming scheme. (b) The energy band diagram of CHEI and HHI programming scheme.

Fig. 3. (Color online) The architecture of (a) ResNet 50 and (b) VGG 16 convolutional neural network.

Fig. 4. (Color online) (a) The proposed scheme to improve both endurance and speed by optimizing the operation scheme for NN online training. (b) The comparison of the V_th tuning speed of FN tunneling and the HHI. (c) The high linearity and symmetric potentiation and depression process using the CHEI and the HHI combined methods.

Fig. 5. (Color online) (a) The I–V curves of the programmed/erased state before and after 109 cycles. (b) Enhancements of endurance at lower MW show the trade-off between MW and endurance. (c) SS value and (d) I_off of different MW and cycles compared with the traditionalprogramming scheme, wherein each box contains 15 different memory cells.

Fig. 6. (Color online) (a) Comparisons between the proposed scheme and the traditional scheme. (b) Read disturbance of different states after 10⁹ cycles. (c) Applications in CIFAR-10 using ResNet50 and Vgg16. Even after 10⁹ cycles, ~90% accuracy can be achieved for the CIFAR-10 task.

Table 1. The benchmark of this work and various non-volatile CIM devices.