Toshiba Develops High-Speed <span style='color:red'>Algorithm</span> and Hardware Architecture for Deep Learning Processor
Trade news

Toshiba Develops High-Speed Algorithm and Hardware Architecture for Deep Learning Processor

Toshiba Memory Corporation today announced the development of a high-speed and high-energy-efficiency algorithm and hardware architecture for deep learning processing with less degradations of recognition accuracy. The new processor for deep learning implemented on an FPGA achieves 4 times energy efficiency compared to conventional ones. The advance was announced at IEEE Asian Solid-State Circuits Conference 2018 (A-SSCC 2018) in Taiwan on November 6.Deep learning calculations generally require large amounts of multiply-accumulate (MAC) operations, and it has resulted in issues of long calculation time and large energy consumption. Although techniques reducing the number of bits to represent parameters (bit precision) have been proposed to reduce the total calculation amount, one of proposed algorithm reduces the bit precision down to one or two bit, those techniques cause degraded recognition accuracy.Toshiba Memory developed the new algorithm reducing MAC operations by optimizing the bit precision of MAC operations for individual filters in each layer of a neural network. By using the new algorithm, the MAC operations can be reduced with less degradation of recognition accuracy.Furthermore, Toshiba Memory developed a new hardware architecture, called bit-parallel method, which is suitable for MAC operations with different bit precision. This method divides each various bit precision into a bit one by one and can execute 1-bit operation in numerous MAC units in parallel. It significantly improves utilization efficiency of the MAC units in the processor compared to conventional MAC architectures that execute in series.Toshiba Memory implemented ResNet50, a deep neural network, on an FPGA using the various bit precision and bit-parallel MAC architecture. In the case of image recognition for the image dataset of ImageNet, the above technique supports that both operation time and energy consumption for recognizing image data are reduced to 25 % with less recognition accuracy degradation, compared to conventional method.
Key word：
Release time：2018-11-07 00:00 reading：1171 Continue reading>>
<span style='color:red'>Algorithm</span> Speeds GPU-based AI Training 10x on Big Data Sets
Trade news

Algorithm Speeds GPU-based AI Training 10x on Big Data Sets

　　IBM Zurich researchers have developed a generic artificial-intelligence preprocessing building block for accelerating Big Data machine learning algorithms by at least 10 times over existing methods. The approach, which IBM presented Monday (Dec. 4) at the Neural Information Processing Systems conference (NIPS 2017) in Long Beach, Calif., uses mathematical duality to cherry-pick the items in a Big Data stream that will make a difference, ignoring the rest.　　“Our motivation was how to use hardware accelerators, such as GPUs [graphic processing units] and FPGAs [field-programmable gate arrays], when they do not have enough memory to hold all the data points” for Big Data machine learning, IBM Zurich collaborator Celestine Dünner, co-inventor of the algorithm, told EE Times in advance of the announcement.　　“To the best of our knowledge, we are first to have generic solution with a 10x speedup,” said co-inventor Thomas Parnell, an IBM Zurich mathematician. “Specifically, for traditional, linear machine learning models — which are widely used for data sets that are too big for neural networks to train on — we have implemented the techniques on the best reference schemes and demonstrated a minimum of a 10x speedup.”　　IBM Zurich researcher Martin Jaggi at ?cole Polytechnique Fédérale de Lausanne (EPFL), also contributed to the machine learning preprocessing algorithm.　　For their initial demonstration, the researchers used a single Nvidia Quadro M4000 GPU with 8 gigabytes of memory training on a 30-Gbyte data set of 40,000 photos using a support vector machine (SVM) algorithm that resolves the images into classes for recognition. The SVM algorithm also creates a geometric interpretation of the model learned (unlike neural networks, which cannot justify their conclusions). IBM’s data preprocessing method enabled the algorithm to run in less than a one minute, a tenfold speedup over existing methods using limited-memory training.　　The key to the technique is preprocessing each data point to see if it is the mathematical dual of a point already processed. If it is, then the algorithm just skips it, a process that becomes increasingly frequent as the data set is processed. “We calculate the importance of each data point before it is processed by measuring how big the duality gap is,” Dünner said.　　“If you can fit your problem in the memory space of the accelerator, then running in-memory will achieve even better results,” Parnell told EE Times. “So our results apply only to Big Data problems. Not only will it speed up execution time by 10 times or more, but if you are running in the cloud, you won’t have to pay as much.”　　As Big Data sets grow, such time- and money-saving preprocessing algorithms will become increasingly important, according to IBM. To show that its duality-based algorithm works with arbitrarily large data sets, the company showed an eight-GPU version at NIPS that handles a billion examples of click-through data for web ads.　　The researchers are developing the algorithm further for deployment in IBM’s Cloud. It will be recommended for Big Data sets involving social media, online marketing, targeted advertising, finding patterns in telecom data, and fraud detection.　　For details, read Efficient Use of Limited-Memory Accelerators for Linear Learning on Heterogeneous Systems, by Dünner, Parnell, and Jaggi.
Key word：
Release time：2017-12-06 00:00 reading：1098 Continue reading>>
Big Data <span style='color:red'>Algorithm</span>s, Languages Expand
Trade news

Big Data Algorithms, Languages Expand

　　The buzz around big data is spawning new algorithms, programming languages, and techniques at the speed of software.　　“Neural networks have been around for a long time. What’s new is the large amounts of data we have to run against them and the intensity of engineering around them,” said Inderpal Bhandari, a veteran computer scientist who was named IBM’s first chief data officer.　　He described work using generative adversarial networks to pit two neural nets against each other to create a better one. “This is an engineering idea that leads to more algorithms — there is a lot of that kind of engineering around neural networks now.”　　In some ways, the algorithms are anticipating tomorrow’s hardware. For example, quantum algorithms are becoming hot because they “allow you to do some of what quantum computers would do if they were available, and these algorithms are coming of age,” said Anthony Scriffignano, chief data scientist for Dun & Bradstreet.　　Deep belief networks are another hot emerging approach. Scriffignano describes it as “a non-regressive way to modify your goals and objectives while you are still learning — as such, it has characteristics of tomorrow’s neuromorphic computers,” systems geared to mimic the human brain.　　At Stanford, the DeepDive algorithms developed by Chris Ré have been getting traction. They help computers understand and use unstructured data like text, tables, and charts as easily as relational databases or spreadsheets, said Stephen Eglash, who heads the university’s data science initiative.　　“Much of existing data is un- or semi-structured. For example, we can read a datasheet with ease, but it’s hard for a computer to make sense of it.”　　So far, Deep Dive has helped oncologists use computers to interpret photos of tumors. It’s being used by the New York attorney general as a law enforcement tool. It’s also in use across a large number of companies working in different domains.　　DeepDive is unique in part because “it IDs and labels everything and then uses learning engines and probabilistic techniques to figure out what they mean,” said Eglash.　　While successful, the approach is just one of many algorithm efforts in academia these days. Others focus on areas such as computer vision or try to ID anomalies in real-time data streams. “We could go on and on,” said Eglash.
Key word：
Release time：2017-06-09 00:00 reading：1185 Continue reading>>

Turn to

/ 1

Popular categories
Semiconductors Circuit Protection Electromechanical Embedded Solutions Connectors, Interconnects Sensors, Transducers Optoelectronics Development Kits Test and Measurement Power Fans, Thermal Management Boxes, Enclosures, Racks Cables, Wires Passive Components Others
  • Week of hot material
  • Material in short supply seckilling
model brand Quote
BD71847AMWV-E2 ROHM Semiconductor
RB751G-40T2R ROHM Semiconductor
MC33074DR2G onsemi
TL431ACLPR Texas Instruments
CDZVT2R20B ROHM Semiconductor
model brand To snap up
TPS63050YFFR Texas Instruments
BU33JA2MNVX-CTL ROHM Semiconductor
STM32F429IGT6 STMicroelectronics
ESR03EZPJ151 ROHM Semiconductor
BP3621 ROHM Semiconductor
IPZ40N04S5L4R8ATMA1 Infineon Technologies
Hot labels
ROHM
IC
Averlogic
Intel
Samsung
IoT
AI
Sensor
Chip
Original authorized brand
More Licensed Brands>>
Information leaderboard
  • Week of ranking
  • Month ranking

About us

Qr code of ameya360 official account

Identify TWO-DIMENSIONAL code, you can pay attention to

AMEYA360 mall (www.ameya360.com) was launched in 2011. Now there are more than 3,500 high-quality suppliers, including 6 million product model data, and more than 1 million component stocks for purchase. Products cover MCU+ memory + power chip +IGBT+MOS tube + op amp + RF Bluetooth + sensor + resistor capacitance inductor + connector and other fields. main business of platform covers spot sales of electronic components, BOM distribution and product supporting materials, providing one-stop purchasing and sales services for our customers.

Please enter the verification code in the image below：

verification code