4:05 PM - 4:35 PM
[17p-E302-6] Frontiers of AI Computing with Oxide Semiconductors
Keywords:oxide semiconductor, AI computing, neuromorphic system
Oxide semiconductors have outstanding advantages, such as, versatile characteristics due to multinary compound, suitability for 3D structure due to feature deposition, easy manufacturing due to amorphous phase, and low cost due to common materials, and have been used in flat-panel displays. The next killer applications to take these advantages are neuromorphic systems, where astronomical high integration is indispensable although rapid speed and parts robustness is not needed so much, which is appropriate for oxide semiconductors. On the other hand, analog computing requires the co-design of materials, fabrication processes, device structures, circuit configurations, computing architectures, systems, and application. In this lecture, we will introduce various researches from several organizations and report our recent results, especially, thin-film memdevices in 3D structures, memristor + capacitor = dynamic memcapacitors with high dynamic range and ultra low power, extended spiking computing principle, local autonomous learning, etc.