• Final Grade: 98/110 • Graduationthesis: LATENTREPLAY-BASEDON-DEVICECONTINUALLEARN- ING USING TRANSFORMERS ON EDGE ULTRA-LOW-POWER IOT PLATFORMS - The thesis explored PULP deployment solutions for continual learn- ing (CL) methods applied on Visual Transformers (ViT) by extending the PULP TrainLib library to support ViT, and by experimenting with latent replay and associated CL methods on transformers.它是用Python和C编写的,它使用了Pytorch库。•课程工作:AI的体系结构和平台;图像处理和计算机视觉;自然语言处理;机器学习和深度学习;组合,决策和优化;乐趣 - aiinindustry; AI的语言和算法; AI的统计和数学方法。
主要关键词