News

tinyML Summit 2019: Industry Leaders Discuss the Future of Power ML Technologies

February 03, 2020 by Luke James

Following the success of the inaugural tinyML Summit 2019, the show is returning this year for its second run on February 12-13 in San Jose, California.

After tinyML 2019’s success, tinyML 2020 will be returning again this year and invites low power machine learning experts—from industry, academia, government, and start-ups all across the world—to join in to share the latest and greatest in the field and to collectively drive the whole ecosystem forward. 

On February 12-13, researchers and experts will gather from Google, Samsung, Microsoft, and Qualcomm, as well as several universities, to come together in San Jose, California and discuss the latest challenges in bringing machine learning to the edge of the network. The hot topic this year: microprocessors running on sensors and battery-powered devices. 

 

What is tinyML Summit?

Tiny machine learning is a fast-growing field of machine learning technologies and applications that include hardware, algorithms, and software capable of performing on-device sensor analytics at very low power, typically in the mW range and lower. This enables a variety of always-on use-cases and targets devices operated by batteries. 

The tinyML (machine learning) summit has been established with the core goal of figuring out how to run machine learning algorithms on the tiniest of microprocessors. At the edge, machine learning will be able to drive better privacy practices, bring down energy consumption, and build novel applications in future devices.

Last year, the tinyML summit revealed that:

  1. Progress on networks, models, and algorithms down to 100kB and below; 
  2. Initial low power applications in audio and vision; and 
  3. Tiny machine learning-capable hardware is becoming good enough for many commercial applications and new architectures.

Although many companies are focused on building specialized silicon for machine learning to train networks inside data centers, the goal of the tinyML Summit and wider community is to bring inference to the smallest of processors, such as an 8-bit microcontroller inside a remote sensor. 

 

tinyML event information.

tinyML Summit event information, including the location and dates of the show. Image Credit: tinyML.

 

tinyML Industry Research Contributions 

There is more to tinyML than its annual Summit. The wider tinyML community is conducting research and experiments in a range of key areas. 

As an idea, one major area where tinyML researchers are currently experimenting is with better data classification by using machine learning on battery-powered edge devices. Latent AI’s CEO, Jags Kandasamy, says that his company is currently in talks with companies that are building augmented- and virtual-reality headsets that want to take their masses of data and classify images seen on the devices. This will ensure only useful data is sent to the cloud for subsequent training. 

On-device classification could massively reduce the amount of data gathered and sent to the cloud, saving on resources such as bandwidth and electricity, the latter being a resource that machine learning and training requires a lot of. 

 

What to Expect from tinyML Summit 2020

This year, tinyML Summit will continue on what was established last year – high-quality talks from guest speakers, poster and demo presentations, open and stimulating discussions, and significant networking opportunities. The scope of the tinyML Summit is to cover the full stack of technologies, from systems to algorithms and software applications, at deep technical levels. This is a unique feature of the summit, say its organizers. 

Just like last year, the majority of participants and guest speakers will come from industry and leading academic research projects. However, this year, particular attention will be given to the most recent progress on algorithm development and the applications and use-cases for tiny machine learning. 

tinyML Summit 2020 will be organized into four technical sessions – Hardware, Systems, Algorithms, and Software – and there will be around 20 presentations from invited guests alongside poster sessions and demonstrations from leading machine learning companies and sponsors.