The Infona portal uses cookies, i.e. strings of text saved by a browser on the user's device. The portal can access those files and use them to remember the user's data, such as their chosen settings (screen view, interface language, etc.), or their login data. By using the Infona portal the user accepts automatic saving and using this information for portal operation purposes. More information on the subject can be found in the Privacy Policy and Terms of Service. By closing this window the user confirms that they have read the information on cookie usage, and they accept the privacy policy and the way cookies are used by the portal. You can change the cookie settings in your browser.
We have succeeded in utilizing In–Ga–Zn–O (IGZO) thin-film devices as synapse elements in a neural network. The electrical conductance is regarded as the connection strength, and the continuous change by flowing electrical current is employed as the connection plasticity based on the modified Hebbian learning as a learning rule. We developed a cellular neural network using the IGZO thin-film devices...
Neural networks are information processing models based on the human brain, and they have been activity studied. However, in order to realize the hardware of the neural network, it is necessary to achieve high integration. In this study, we fabricated synapses for neural networks using the Ga-Sn-O (GTO) thin films, which is rare-metal-free amorphous oxide semiconductor. The synapses are planar type,...
Nuromorphic devices are powerful information processing systems and realize low power consumption. We proposed cellular neural networks with polycrystalline-silicon (Poly-Si) thin film transistors (TFTs). The neuron element consists of eight TFTs and the synapse element of only one TFT. We actually succeeded in confirming the operation of learning of logics. In this presentation, we succeeded the...
We have developed a cross-point device using In-Ga-Zn-O (IGZO) semiconductor for synapse element in neural network. Horizontal 80 and vertical 80 metal lines make 6400 cross-point synapse integrated on a glass substrate. The electrical conductance gradually degrades by flowing current, which is available for modified Hebbian learning.
Neural networks are computing models based on human brains. We propose a neural network using a field-programmable gate array (FPGA) for neurons and amorphous In-Ga-Zn-O (a-IGZO) thin films for synapses. It is found that electric current in the a-IGZO thin film gradually decreases along the time. On the other hand, the degradation does not occur when light is irradiated. These phenomena can be utilized...
We are developing cellular neural networks using thin-film transistors (TFTs). Although simplification of the circuits for the neurons and synapses is also needed for the cellular neural network, the detailed discussion is not sufficient for the cellular neural network. Particularly in this study, we tried simplification of synapse devices. We used discrete trimmer resistors and capacitors for the...
We are developing neural networks using thin-film transistors (TFTs). By adopting an interconnect-type neural network and utilizing a characteristic degradation of poly-Si TFTs as a variable strength of synapse connection, which was originally an issue, we realized the neuron consisting of eight TFTs and synapse of only one TFT. Particularly in this presentation, we confirmed that the learning efficiency...
We are developing cellular neural networks using thin-film transistors (TFTs). Although simplification of the processing elements such as neurons and synapses is also needed for the cellular neural network, it is difficult and time-consuming to fabricate and evaluate actual devices. Therefore, we are studying the simplification of the processing elements in the neural networks by using circuit simulation...
We have developed neuron MOS devices using TFTs. We fabricated and evaluated neuron MOS inverter, which can be also utilized as a variable threshold voltage inverter, and neuron MOS source follower, which can be also utilized as a digital-analog converter. The neuron MOS devices indicate that TFTs have great potential for application to artificial neural networks.
Set the date range to filter the displayed results. You can set a starting date, ending date or both. You can enter the dates manually or choose them from the calendar.