An Enhanced Floating Gate Memory for the Online Training of Analog Neural Networks

Floating gate (FG) memory has long erasing time, which limits its application as an electronic synapse in online training. This paper proposes a novel enhanced floating gate memory (EFM) by TCAD simulation. Here, three other structures are simulated just for comparison. The simulation results show that the erasing speed is about 34ns while the other three need the time over 1.8ms, which makes the operation speed of long-term potentiation (LTP) more symmetrical to long-term depression (LTD). In addition, both LTP and LTD are approximately linear in the simulation results. The speed, linearity, and symmetry of weight update are the keys to online training of analog neural networks. These excellent performances indicated a potential application of EFM in analog neuro-inspired computing.

[1]  V. Hu,et al.  EEPROM device as a reconfigurable analog element for neural networks , 1989, International Technical Digest on Electron Devices Meeting.

[2]  Shimeng Yu,et al.  Mitigating effects of non-ideal synaptic device characteristics for on-chip learning , 2015, 2015 IEEE/ACM International Conference on Computer-Aided Design (ICCAD).

[3]  Shimeng Yu,et al.  Neuro-Inspired Computing With Emerging Nonvolatile Memorys , 2018, Proceedings of the IEEE.

[4]  Tadashi Shibata,et al.  An excellent weight-updating-linearity EEPROM synapse memory cell for self-learning Neuron-MOS neural networks , 1993 .

[5]  Xinjie Guo,et al.  Redesigning commercial floating-gate memory for analog computing applications , 2014, 2015 IEEE International Symposium on Circuits and Systems (ISCAS).

[6]  H. Zhu,et al.  A Semi-Floating Gate Transistor With Enhanced Embedded Tunneling Field-Effect Transistor , 2018, IEEE Electron Device Letters.

[7]  Peng Zhou,et al.  Flexible Electronic Synapses for Face Recognition Application with Multimodulated Conductance States. , 2018, ACS applied materials & interfaces.

[8]  Wei Wang,et al.  Design of U-Shape Channel Tunnel FETs With SiGe Source Regions , 2014, IEEE Transactions on Electron Devices.

[9]  Jang-Sik Lee,et al.  Short-Term Plasticity and Long-Term Potentiation in Artificial Biosynapses with Diffusive Dynamics. , 2018, ACS nano.

[10]  Shimeng Yu,et al.  Fully parallel write/read in resistive synaptic array for accelerating on-chip learning , 2015, Nanotechnology.

[11]  F. Merrikh Bayat,et al.  Model-based high-precision tuning of NOR flash memory cells for analog computing applications , 2016, 2016 74th Annual Device Research Conference (DRC).

[12]  Youngjune Park,et al.  Artificial Synapses with Short- and Long-Term Memory for Spiking Neural Networks Based on Renewable Materials. , 2017, ACS nano.

[13]  Dharmendra S. Modha,et al.  Backpropagation for Energy-Efficient Neuromorphic Computing , 2015, NIPS.

[14]  S. Tam,et al.  An electrically trainable artificial neural network (ETANN) with 10240 'floating gate' synapses , 1990, International 1989 Joint Conference on Neural Networks.

[15]  R. C. Hughes.  High field electronic properties of SiO2 , 1978 .

[16]  Pritish Narayanan,et al.  Experimental Demonstration and Tolerancing of a Large-Scale Neural Network (165 000 Synapses) Using Phase-Change Memory as the Synaptic Weight Element , 2014, IEEE Transactions on Electron Devices.

[17]  S. Loth,et al.  A Logical Use for Atoms , 2011, Science.

[18]  F. Merrikh Bayat,et al.  Fast, energy-efficient, robust, and reproducible mixed-signal neuromorphic classifier based on embedded NOR flash memory technology , 2017, 2017 IEEE International Electron Devices Meeting (IEDM).

[19]  Farnood Merrikh-Bayat,et al.  Temperature-insensitive analog vector-by-matrix multiplier based on 55 nm NOR flash memory cells , 2016, 2017 IEEE Custom Integrated Circuits Conference (CICC).