Efficient and Reliable Deep Learning

高效可靠的深度学习

• 批准号: RGPIN-2021-04012
• 负责人: Li, Ke
• 金额: $ 2.48万
• 依托单位: Simon Fraser University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=750566
• 关键词: Efficient; Reliable; Deep; Learning

Machine learning is the embodiment of an unapologetically data-driven philosophy that has increasingly become one of the most important drivers of progress in artificial intelligence (AI) and beyond. In recent years, exciting advances in deep learning have revolutionized the field and paved the way for applications that were considered infeasible as recently as a decade ago, from speech recognition to image synthesis. The potential of deep learning has not yet been realized to its fullest however, due to challenges along two dimensions: reliability and efficiency. The former is necessary to bring deep learning to real world applications involving consequential decisions, and the latter is necessary to broaden the applicability of deep learning to low-resource settings. We propose research that advances deep learning along these dimensions by developing methods that (1) know when they could fail and (2) can be trained and evaluated efficiently.

机器学习毫无疑问是数据驱动哲学的体现，这种哲学已日益成为人工智能 (AI) 及其他领域进步的最重要驱动力之一。近年来，深度学习令人兴奋的进步彻底改变了该领域，并为十年前被认为不可行的应用（从语音识别到图像合成）铺平了道路。 然而，由于可靠性和效率两个方面的挑战，深度学习的潜力尚未得到充分发挥。前者是将深度学习引入涉及后续决策的现实世界应用所必需的，后者是将深度学习的适用性扩大到资源匮乏的环境所必需的。我们提出研究，通过开发（1）知道何时可能失败以及（2）可以有效地训练和评估的方法来推进深度学习在这些方面的发展。