Malay Haldar, Mustafa Abdool, Prashant Ramanathan, Tao Xu, Shulin Yang, Huizhong Duan, Qing Zhang, Nick Barrow-Williams, Bradley C. Turnbull, Brendan M. Collins, Thomas Legrand
The application to search ranking is one of the biggest machine learning success stories at Airbnb. Much of the initial gains were driven by a gradient boosted decision tree model. The gains, however, plateaued over time. This paper discusses the work done in applying neural networks in an attempt to break out of that plateau. We present our perspective not with the intention of pushing the frontier of new modeling techniques. Instead, ours is a story of the elements we found useful in applying neural networks to a real life product. Deep learning was steep learning for us. To other teams embarking on similar journeys, we hope an account of our struggles and triumphs will provide some useful pointers. Bon voyage!
DeepPicarMicro: Applying TinyML to Autonomous Cyber Physical Systems
University of Kansas（堪萨斯大学）
在微型微控制器单元（MCU）上运行深度神经网络（DNN）具有挑战性，因为它们在计算、内存和存储容量方面存在局限性。幸运的是，MCU硬件和机器学习软件框架的最新进展使得在现代MCU上运行相当复杂的神经网络成为可能，从而产生了一个被广泛称为TinyML的新研究领域。然而，很少有研究表明TinyML在网络物理系统（CPS）中的应用潜力。在本文中，我们介绍了DeepPicarMicro，一个小型自动驾驶RC汽车测试平台，它在Raspberry Pi Pico MCU上运行卷积神经网络（CNN）。我们采用了最先进的DNN优化技术，成功地将著名的PilotNet CNN架构（用于驱动NVIDIA的真正自动驾驶汽车）安装在MCU上。我们应用最先进的网络架构搜索（NAS）方法来寻找进一步优化的网络，这些网络可以以端到端的方式实时有效地控制汽车。从一个广泛的系统的实验评估研究，我们观察到一个有趣的准确性，延迟和系统的控制性能之间的关系。由此，我们提出了一种联合优化策略，该策略在启用AI的CPS的网络架构搜索过程中同时考虑模型的准确性和延迟。
