New chip reduces neural networks’ power consumption by up to 95 percent, making them practical for battery-powered devices.
Most recent advances in artificial-intelligence systems such as speech- or face-recognition programs have come courtesy of neural networks, densely interconnected meshes of simple information processors that learn to perform tasks by analyzing huge sets of training data.
But neural nets are large, and their computations are energy intensive, so they’re not very practical for handheld devices. Most smartphone apps that rely on neural nets simply upload data to internet servers, which process it and send the results back to the phone.
Now, MIT researchers have developed a special-purpose chip that increases the speed of neural-network computations by three to seven times over its predecessors, while reducing power consumption 94 to 95 percent. That could make it practical to run neural networks locally on smartphones or even to embed them in household appliances.
“The general processor model is that there is a memory in some part of the chip, and there is a processor in another part of the chip, and you move the data back and forth between them when you do these computations,” says Avishek Biswas, an MIT graduate student in electrical engineering and computer science, who led the new chip’s development.
“Since these machine-learning algorithms need so many computations, this transferring back and forth of data is the dominant portion of the energy consumption. But the computation these algorithms do can be simplified to one specific operation, called the dot product. Our approach was, can we implement this dot-product functionality inside the memory so that you don’t need to transfer this data back and forth?”
Biswas and his thesis advisor, Anantha Chandrakasan, dean of MIT’s School of Engineering and the Vannevar Bush Professor of Electrical Engineering and Computer Science, describe the new chip in a paper that Biswas is presenting this week at the International Solid State Circuits Conference.
Back to analog
Neural networks are typically arranged into layers. A single processing node in one layer of the network will generally receive data from several nodes in the layer below and pass data to several nodes in the layer above. Each connection between nodes has its own “weight,” which indicates how large a role the output of one node will play in the computation performed by the next. Training the network is a matter of setting those weights.
A node receiving data from multiple nodes in the layer below will multiply each input by the weight of the corresponding connection and sum the results. That operation — the summation of multiplications — is the definition of a dot product. If the dot product exceeds some threshold value, the node will transmit it to nodes in the next layer, over connections with their own weights.
A neural net is an abstraction: The “nodes” are just weights stored in a computer’s memory. Calculating a dot product usually involves fetching a weight from memory, fetching the associated data item, multiplying the two, storing the result somewhere, and then repeating the operation for every input to a node. Given that a neural net will have thousands or even millions of nodes, that’s a lot of data to move around.
But that sequence of operations is just a digital approximation of what happens in the brain, where signals traveling along multiple neurons meet at a “synapse,” or a gap between bundles of neurons. The neurons’ firing rates and the electrochemical signals that cross the synapse correspond to the data values and weights. The MIT researchers’ new chip improves efficiency by replicating the brain more faithfully.
In the chip, a node’s input values are converted into electrical voltages and then multiplied by the appropriate weights. Summing the products is simply a matter of combining the voltages. Only the combined voltages are converted back into a digital representation and stored for further processing.
The chip can thus calculate dot products for multiple nodes — 16 at a time, in the prototype — in a single step, instead of shuttling between a processor and memory for every computation.
All or nothing
One of the keys to the system is that all the weights are either 1 or -1. That means that they can be implemented within the memory itself as simple switches that either close a circuit or leave it open. Recent theoretical work suggests that neural nets trained with only two weights should lose little accuracy — somewhere between 1 and 2 percent.
Biswas and Chandrakasan’s research bears that prediction out. In experiments, they ran the full implementation of a neural network on a conventional computer and the binary-weight equivalent on their chip. Their chip’s results were generally within 2 to 3 percent of the conventional network’s.
“This is a promising real-world demonstration of SRAM-based in-memory analog computing for deep-learning applications,” says Dario Gil, vice president of artificial intelligence at IBM. “The results show impressive specifications for the energy-efficient implementation of convolution operations with memory arrays. It certainly will open the possibility to employ more complex convolutional neural networks for image and video classifications in IoT [the internet of things] in the future.”
Learn more: Neural networks everywhere
The Latest on: Neural networks
- Malware Detection With Convolutional Neural Networks in Python on September 19, 2018 at 9:32 am
Insight for I&O leaders on deploying AIOps platforms to enhance performance monitoring today. Read the Guide. In this post, we will learn about artificial network architectures and how to use one ... […]
- This Lion Eats Words, Roars Poetry, Is Powered by Google’s Neural Network on September 19, 2018 at 2:19 am
The lion uses a long short-term memory (LSTM) recurrent neural network (RNN) to predict linear sequences. It was inspired by Google’s efforts to train deep learning models to write Visitors to Trafalg... […]
- Scientists use artificial neural networks to predict new stable materials on September 18, 2018 at 3:05 pm
We allow third-party companies to serve ads and/or collect anonymous information. These companies may use non-personally identifiable information (browser type, time and date) in order to provide adve... […]
- DarwinAI raises $3 million for AI that optimizes neural networks on September 18, 2018 at 7:55 am
Machine learning is a powerful tool, no doubt, capable of diagnosing lung cancer, besting champion Go players, and navigating the labyrinthine streets of San Francisco. But the neural networks ... […]
- Toward Self-Creating Neural Networks That Explain Themselves on September 18, 2018 at 4:09 am
On the hardware side, the next frontier for deep learning innovation will be in getting the performance, efficiency, and accuracy needed for inference at scale. But the newest battleground in software ... […]
- Artificial neural network now capable of finding medication complaints in social networks on September 17, 2018 at 9:28 pm
Graphical representation of semantic text analysis in a recurrent neural network Credit: Kazan Federal University A joint paper by Kazan Federal University and Moscow Institute of Physics and ... […]
- Machine learning? Neural networks? Here’s your guide to the many flavors of A.I. on September 15, 2018 at 2:39 am
A.I. is everywhere at the moment, and it’s responsible for everything from the virtual assistants on our smartphones to the self-driving cars soon to be filling our roads to the cutting-edge image rec... […]
- MIT taught a neural network how to show its work on September 12, 2018 at 12:44 pm
MIT’s Lincoln Laboratory Intelligence and Decision Technologies Group yesterday unveiled a neural network capable of explaining its reasoning. It’s the latest attack on the black box problem ... […]
- First Wave of Spiking Neural Network Hardware Hits on September 11, 2018 at 3:59 pm
Over the last several years we have seen many new hardware architectures emerge for deep learning training but this year, inference will have its turn in the spotlight. For those chips that can manage ... […]
- SETI's New Neural Network Detects Many More Fast Radio Bursts From a Galaxy Far, Far Away on September 10, 2018 at 8:10 pm
An artist’s depiction of a rupturing magnetar, a rotating neutron star with an extreme magnetic field that scientists suspect may be one origin of fast radio bursts. Illustration: NASA’s ... […]
via Google News and Bing News