Embedded AI localizes data processing for greater speed and security | Research & Technology | November 2022

Dresden, Germany, November 17, 2022 – The Fraunhofer Institute for Photonic Microsystems (IPMS) aims to support safer and faster data processing by integrating machine learning algorithms into digital devices.

Although AI-enabled devices are tightly integrated with daily life, processing of data inputs takes place on large external servers. Included Artificial intelligence Edge AI is set to change this by allowing these processing tasks to be done directly on the device. However, AI performance, especially on very small devices, has so far been limited.


Fraunhofer IPMS integrates artificial intelligence into microsensors and actuators. Courtesy of Fraunhofer IPMS.


The researchers at Fraunhofer IPMS are working to remedy this through networking experiences and developments from disparate areas of research. For example, in an internal institute project, findings from microsensor and actuator technology were combined with the latest in nanoelectronics, wireless communications, and processor developments.

The combination enables pre-processing of signals associated with a sensor or actuator using AI-based methods, providing advantages in lower latency processing and more secure data processing while avoiding the need for network connectivity. In addition, the use of AI edge data processing will enable local re-learning in the field, so that the system can be optimized for specific on-site conditions.

Fraunhofer IPMS's Li-Fi GigaDock enables optical data transmission of large amounts of data at low latency.  Courtesy of Fraunhofer IPMS.


Fraunhofer IPMS’s Li-Fi GigaDock enables optical data transmission of large amounts of data at low latency. Courtesy of Fraunhofer IPMS.


Applications of this technology include spectrometers, ISFET (Ion Sensitive Field Effect Transistor) sensors, ultrasound imaging for state monitoring, gesture control, or environment recognition for collaborative robots.

To support the integration of edge AI technology across sensor and actuator technologies, the researchers extended the existing EMSA5 RISC-V computing platform, with Tensorflow Lite-based AI functionality. Fraunhofer showed the setup at the Electronica trade fair in Munich on November 15-18. The researchers also presented some of the institute’s latest developments in the field of core technology for intellectual property and optical wireless data transmission in electronica.

.

Source

IBM gives partners access to three new AI libraries

The new libraries provide natural language processing, speech-to-text, and text-to-speech capabilities for partners to add to applications for any hybrid and multi-cloud environment, says Monica Agarwal, IBM’s VP of Ecosystem Building and Technology Partnerships.

IBM has released new AI libraries designed to help IBM partners, customers and developers build and bring AI-powered services to market.

The three new AI libraries for New York-based Armonk, now available publicly, combine innovation from IBM Research and open source technology to provide Natural Language Processing, Speech-to-Text and Text-to-Speech capabilities for partners to add to the new, said Monica Aggarwal, Vice President Head of ISV Build Ecosystem and Technology Partnerships at IBM for CRN, Applications for any hybrid and multi-cloud environment.

Aggarwal said this expanded access to AI will help MSPs, MSSPs and other similar partners to accelerate AI adoption.

“We got feedback from partners, we listened to them, we worked on our strategy and we said, ‘You know what? Let’s build something very convenient for consumption, said Aggarwal. “A simple, easy and flexible format for libraries to embed in a solution. And we have a great program to build with you, go to market with you – let’s take that and win the market.”

 

What has IBM done for artificial intelligence?

Nalit Patel, CEO of All Solutions, an IBM partner in Livingston, New Jersey, told CRN that he loves IBM’s investment in cutting-edge technology, including artificial intelligence.

He also said he would love to invest more in the IBM Partner Program.

They are becoming more channel friendly,” Patel said.

Meanwhile, the libraries are giving partners access to the same AI libraries that power IBM Watson products, according to IBM. Developers and IT teams can include new libraries in their applications and create custom products without data science expertise, reducing the barrier to AI adoption.

One library, the IBM Watson Natural Language Processing Library, is used to process human language and discover meaning, content, intent and feelings, according to the vendor.

The IBM Watson Speech to Text Library is used for speech transcription for customer service. The IBM Watson Text to Speech library is used to convert typed text to natural voice in a variety of languages ​​and sounds in an application.

Earlier this month, during IBM’s latest quarterly earnings call, IBM Chairman and CEO Arvind Krishna Call Artificial intelligence and hybrid cloud are “the two most transformative enterprise technologies of our time” and described partners with different business models as a “critical component of our strategy” to win customers.

IBM has also been in a series of acquisitions to build its AI capabilities along with innovation from within the company. Recent AI-related acquisitions include Databand.ai And the wig And the new.

 

Source

Snap Way to Design Ad Ranking Service Using Deep Learning

Snap geometry recently posted a file Blog Posts on how they designed their ad categorization and targeting service using deep learning.

Showing ads to users is the trend of monetizing social networking platform. Snap’s ad rating system is designed to target the right user at the right time. Snap follows a few principles in designing such a complex system. They focus on providing an excellent user experience while maintaining user privacy and security. The following image shows the high-level structure of Snap Ads Ranking and Targeting.

Snap .’s Comprehensive Ad Classification Service Architecture

This system consists of some microservices. It contains a module called Ad Eligibility Filter. This module applies some filters such as checking ad budget, privacy, and ad policy compliance. Determines whether an advertisement is appropriate to display to the user. Ads from this module are passed to the candidate generation module. The filter generation module identifies relevant ads to users. Usually a machine learning model is very light Re-Call used in this unit. Once hundreds or a few thousand ads are identified, a heavy deep learning model is used to score the likelihood of the ads converting. In the end, the auction is run to select and score ads based on business rules, budget, etc. The feedback loop is used to collect more user interaction data to retrain models.

Ad rating systems face some unique challenges compared to general recommendation and search ranking systems. First of all, size, cost and latency. As mentioned in Blog Posts :

Our ML models work very broadly; Every day we make trillions of predictions using models trained on billions of examples. Such a large scale involves significant training and heuristics costs. These models also operate under strict latency restrictions. A combination of highly optimized training platform, inference platform, and model architectures is needed to keep cost and latency within acceptable limits.

Low prediction error rate is an important factor in choosing an advertisement. The error in predicting the outcome will propagate in the auction phase resulting in low quality data in the model’s training. The rapid change in advertising inventory is also one of the main challenges. This makes it difficult to get good embed Representation of advertisements.

For heavy ml models, in-service deep learning models are used as described in Blog Posts :

We make use of the latest multitasking models like MMoEs And the PLE To jointly predict multiple conversion events (eg, app installs, purchases, and subscriptions). Our models also use the latest high-level feature interaction layers such as DCN And the DCN version 2 as a building stone. We also optimize these models for inference cost and response time by splitting them into multiple towers, ie one for user features processing and one for advertising features. We illustrate these components and their interrelationships in the figure below.

A typical high-level structure used to classify Snap ads

Ad Ranking Systems is a use case recovery systems. Some cloud service providers such as AmazonAnd the The GoogleAnd the Microsoft It also offers these return systems as a service to its customers. There are some open source systems like Ad search systems On Github for further review.

.

Source

Exit mobile version