A5008795949- Advanced Memory and Neural Computing
- Neural dynamics and brain function
- Neuroscience and Neural Engineering
- Ferroelectric and Negative Capacitance Devices
- Neural Networks and Reservoir Computing
- Neural Networks and Applications
- CCD and CMOS Imaging Sensors
- Advanced Chemical Sensor Technologies
- Gas Sensing Nanomaterials and Sensors
- Semiconductor materials and devices
- Quantum-Dot Cellular Automata
- Analytical Chemistry and Sensors
- Insect Pheromone Research and Control
Seoul National University
2018-2024
Kookmin University
2020
In this paper, we reviewed the recent trends on neuromorphic computing using emerging memory technologies. Two representative learning algorithms used to implement a hardware-based neural network are described as bio-inspired algorithm and software-based algorithm, in particular back-propagation. The requirements of synaptic device apply each were analyzed. Then, research devices an artificial network.
Hardware-based spiking neural networks (SNNs) inspired by a biological nervous system are regarded as an innovative computing with very low power consumption and massively parallel operation. To train SNNs supervision, we propose efficient on-chip training scheme approximating backpropagation algorithm suitable for hardware implementation. We show that the accuracy of proposed is close to conventional artificial (ANNs) using stochastic characteristics neurons. In configuration, gated...
Abstract Smart healthcare systems integrated with advanced deep neural networks enable real‐time health monitoring, early disease detection, and personalized treatment. In this work, a novel 3D AND‐type flash memory array rounded double channel for computing‐in‐memory (CIM) architecture to overcome the limitations of conventional smart systems: necessity high area energy efficiency while maintaining classification accuracy is proposed. The fabricated array, characterized by low‐power...
Hardware-based spiking neural networks (SNNs) are regarded as promising candidates for the cognitive computing system due to low power consumption and highly parallel operation. In this work, we train SNN in which firing time carries information using temporal backpropagation. The temporally encoded with 512 hidden neurons showed an accuracy of 96.90% MNIST test set. Furthermore, effect device variation on is investigated compared that rate-encoded network. a hardware configuration our SNN,...
Spiking neural networks (SNNs) have emerged as a novel approach for reducing computational costs by mimicking the biologically plausible operations of neurons and synapses. In this article, large‐scale analog SNNs are investigated optimized at hardware‐level using SNNSim, simulator that employ synaptic devices integrate‐and‐fire (I&F) neuron circuits. SNNSim is reconfigurable accurately very quickly models behavior user‐defined device characteristics returns key metrics such area,...
Brain-inspired analog neuromorphic systems based on the synaptic arrays have attracted large attention due to low-power computing. Spike-timing-dependent plasticity (STDP) algorithm is considered as one of appropriate neuro-inspired techniques be applied for on-chip learning. The aim this study investigate methodology unsupervised STDP learning in temporal encoding systems. system-level simulation was performed measurement results thin-film transistor-type asymmetric floating-gate NOR flash...
Abstract Reinforcement learning (RL) using deep Q-networks (DQNs) has shown performance beyond the human level in a number of complex problems. In addition, many studies have focused on bio-inspired hardware-based spiking neural networks (SNNs) given capabilities these technologies to realize both parallel operation and low power consumption. Here, we propose an on-chip training method for DQNs applicable SNNs. Because conventional backpropagation (BP) algorithm is approximated, evaluation...
In hardware-based spiking neural networks (SNNs), the conversion of analog input data into arrival time an pulse is regarded as a good candidate for encoding method due to its bio-plausibility and power-efficiency. this work, we trained SNN encoded by first spike (TTFS) performed inference process using behavior fabricated TFT-type flash synaptic device. The exponentially decaying current model required in was implemented reading devices subthreshold region triangle pulses. high-level system...
We present a two-layer fully connected neuromorphic system based on thin-film transistor (TFT)-type NOR flash memory array with multiple postsynaptic (POST) neurons. Unsupervised online learning by spike-timing-dependent plasticity (STDP) the binary MNIST handwritten datasets is implemented, and its recognition result determined measuring firing rate of POST Using proposed scheme, we investigate impact number neurons in terms rate. In this system, lateral inhibition function homeostatic...
Artificial intelligence technology has attracted much attention in recent years, and technological progress of this is anticipated with the development semiconductor technology. This talk focuses on synaptic mimic devices to realize artificial memory These affect cognitive accuracy along conductance quantization architecture. Therefore, we will first discuss from architectural point view examine characteristics candidates for various synapse being reported. In particular, concentrate...
Artificial Neural Networks (ANNs) have shown remarkable performance in various fields. However, ANN relies on the von-Neumann architecture, which consumes a lot of power. Hardware-based spiking neural networks (SNNs) inspired by human brain become an alternative with significantly low power consumption. In this paper, we propose on-chip trainable SNNs using time-to-first-spike (TTFS) method. We modify learning rules conventional TTFS to be suitable for learning. Vertical NAND flash memory...
As a synaptic device, TFT-type NOR flash memory cell shows reasonable weight levels (50 for long-term potentiation (LTP) and 150 depression (LTD)) large max/min ratio (═50) synapse weight. Based on the measurement results of cell, supervised learning process is simulated using software MATLAB. A new pulse scheme designed mimicking spike-rate-dependent plasticity (SRDP) algorithm. Through inferencing phase, our (784 × 100) network achieved 74.08% accuracy MNIST benchmark. method adapting...
We present a two-layer fully connected neuromorphic system based on thin-film transistor (TFT)-type NOR flash memory array with multiple postsynaptic (POST) neurons. Unsupervised online learning by spike-timing-dependent plasticity (STDP) the binary MNIST handwritten datasets is implemented, and its recognition result determined measuring firing rate of POST Using proposed scheme, we investigate impact number neurons in terms rate. In this system, lateral inhibition function homeostatic...
For more efficient computation, researches on neuromorphic systems suitable for neural networks have been conducted [1]. To replace conventional neuron circuits integrate-and-fire (IF) function in systems, devices with high density and low power consumption per spike studied [2, 3]. In the previous work, an n -type positive feedback (PF) device steep switching characteristic was proposed hardware-based [4]. this paper, a p PF charge storage layer is reported as neurons, which integrates...