China
Africa
Explore chapters and articles related to this topic
Machine Learning Frameworks and Device Engineering
Published in Chandrasekar Vuppalapati, Democratization of Artificial Intelligence for the Future of Humanity, 2021
The Model AI/ML Accuracy can be influenced by various factors of hardware and available connectivity options. The factors that influence model accuracy include: Hardware Refresh—this parameter predicates the rate of refresh of the AI/ML model on the Constrained Device.Over the Air (OTA) Firmware Update: OTA is a direct upgrade of a Firmware on to a target device.Eager Learners vs. Lazy Learners ML Model Option: The Machine Learning Learner that performs model inference based on limited set of data vs. fully backed model. Examples of Eager Learner—Decision Tree, Naïve Bayesian and Artificial Neural Networks (ANN) and examples of lazy learner include K-nearest Neighbor. Due to the model construction, eager learners take a long time to train and less time to predict. Compared to eager learners, lazy learners have less training time but more time in predicting.Data Movement needs of the Model: Model performs that needs huge memory transfer between CPU to Memory.Model Invocation o In-line AI Modelo Stack or Heap Load Models
Selection of Emerging Technologies: A Case Study in Technology Strategies of Intelligent Vehicles
Published in Engineering Management Journal, 2022
Fuquan Zhao, Xu Kuang, Han Hao, Zongwei Liu
Nowadays vehicle controllers have become a new market that attracts enterprises to promote the relevant research and application. Conventional driver assistance systems usually adopt DSP solutions, and incumbent firms in automotive semiconductor industry like NXP, Freescale, and Texas Instruments all possess skills and knowledge in the field of DSP. In contrast, new entrants select technology paths according to their own advantages. For instance, DRIVE PX AI car computers launched by NVIDIA are able to achieve 360° situational awareness, precise positioning, trajectory computation, deep learning, and over-the-air updating, with GPUs as core components (NVIDIA, 2017). And the existing market of GPUs have already been dominated by NVIDIA. Moreover, as the leader of CPU market, Intel released the AD solution Intel Go in combination of CPUs and FPGAs. This product has a flexible and scalable architecture to support 5 G communication and V2X applications. In regard to startups like Mobileye, Horizon Robotics, and Cambricon Technologies, or new players of semiconductor such as Google and Microsoft, FPGA and ASIC are more suitable, because these solutions have lower technological barriers and higher degree of customization.