China
Africa
Explore chapters and articles related to this topic
Performance Analysis of Deep Convolutional Neural Networks for Diagnosing COVID-19: Data to Deployment
Published in K. Gayathri Devi, Kishore Balasubramanian, Le Anh Ngoc, Machine Learning and Deep Learning Techniques for Medical Science, 2022
The model has been trained and validated on Google COLAB platform which is a very useful tool for executing deep learning algorithms as it gives free access to dedicated GPUs and TPUs (Tensor Processing Unit). The base hardware for the training face of the model is the TPU offered by Google over the cloud. A TPU is a hardware accelerator specialized for deep learning tasks. It shortens the training time by performing matrix multiplication in the hardware. TPU is a 65,536 8-bit MAC matrix multiply unit that offers a peak throughput of 92 TeraOps/second (TOPS). According to the authors, the matrix unit uses systolic execution to save energy and time by reducing reads and writes of the buffer. Subsequently, the validation and testing phases were carried out on Tesla K80 GPU over the cloud.
Artificial Intelligence and the Cloud
Published in Frank M. Groom, Stephan S. Jones, Artificial Intelligence and Machine Learning for Business for Non-Engineers, 2019
Large AIaaS providers have also created new services and products. Google is currently beta testing a service they call Cloud TPU (tensor processing unit). A TPU is an AI and machine learning accelerator. It functions in the same way as an application-specific integrated circuit (ASIC). ASIC functions similarly to microchips; however, instead of processing many general-purpose functions, they only process one specific function. This makes the process run much faster than if the processor had to run multiple tasks at one time. This same technology is one technique people use to mine Bitcoin. Instead of using it to mine cryptocurrencies, Google uses them in their AIaaS platform to accelerate machine learning and data training (Techopedia, n.d.). This could make the training of data go from taking a few days to a week. Initially, the technology was available only for some users on the market, but was made available to everyone in the last half of 2018 according to Google. Cloud TPU has the possibility of making one of the most difficult tasks for AI and making it a shorter process, allowing for the companies to have more time to train and run AI services (Osborne, 2016).
High-Performance Computing and Its Requirements in Deep Learning
Published in Sanjay Saxena, Sudip Paul, High-Performance Medical Image Processing, 2022
Biswajit Jena, Gopal Krishna Nayak, Sanjay Saxena
The central processing units (CPUs) have been largely used because of their popularity to solve multiple computing applications hassle-freely. It was initially designed for computing purposes in all kinds of applications involving arithmetic and logic operation. It has been used mostly on laptops, desktops, embedded systems, and mobile devices to run software and application of various ranges. The CPU also supports various kinds of operating systems of 32 bit or 64bit, such as Linux, Windows, Mac OS, and all kinds of real-time operating systems (RTOS). Even if the CPU (Central Processing Unit) can process visual data computing, it still has some limitations. While GPUs (Graphical Processing Unit) have thousands of cores present in them, CPUs have limited. Hence CPUs are good for serial processing, whereas GPUs are preferred for parallel processing. In the case of GPU, various digital signal processing operations are performed to make the addition, subtraction, multiplication, division, and high-end applications like gaming, high-definition video editing, streaming, etc. It also needs some extra hardware to perform these applications. TPU (Tensor Processing Unit) [28–30] is a customized central processor worked by Google to do huge loads of information preparation at low accuracy. It is utilized by numerous individuals of the applications which google offers like google search, google map or google photographs. The best thing about TPU is its Artificial Intelligence features, such as machine learning and deep learning. Likewise, it isn’t ready for selling purposes now, and different TPUs are associated with the structure of a supercomputer to crunch a colossal measure of information that google creates from its server.
Simplified Prediction Method for Detecting the Emergency Braking Intention Using EEG and a CNN Trained with a 2D Matrices Tensor Arrangement
Published in International Journal of Human–Computer Interaction, 2023
Hermes J. Mora, Esteban J. Pino
TPU board is an application-specific integrated circuit (ASIC) that exclusively accelerating the machine learning workload (AI process) (Lu et al., 2020). TPUs are designed to use TensorFlow for high performance and flexibility in research projects. They are highly optimized for large batches in CNNs and have the highest training throughput. By contrast, GPUs have better flexibility to manage irregular computations such as in small batches, achieving a high FLOPS usage for RNNs. A TPU can handle up to 128,000 operations per cycle that are ten times the operations handled by a GPU and 100,000× times the speed of a CPU depending on the model and final configuration (Jouppi et al., 2017; S. Kumar et al., 2021; Yazdanbakhsh et al., 2021). However, this hardware accelerator represents maintenance costs and energy consumption in individualized work platforms.