發展使用高能源與面積效率的記憶體運算之AI加速平台，實現從AI模型到硬體自動化，可用於即時推理與終端訓練，並發展協同式人工神經網路計算平台，不依賴雲端，有效整合眾多AI終端。

This project will develop a high energy and area efficient AI accelerator platform with in-memory computing, which automatically maps a software AI model to hardware platform, and is suitable for real time edge AI inference and training without cloud service. We will develop a brain-inspired in-memory computing architecture to implement AI edge nodes. Also, we will design deep learning algorithm and architecture to fully exploit the parallel processing and high density benefits of in-memory computing circuits and device to avoid high cost multiplier-accumulator.