AI Tech

Model Search by Google automatically optimizes and identifies AI models

Google announced the release of Model Search, an open-source platform designed to help researchers develop ML models efficiently and automatically. Instead of focusing on a specific domain, Google’s Model Search is domain-agnostic, making it capable of finding a model architecture that fits a dataset and problem while minimizing coding time and compute resources.

The success of an AI model often depends on how well it can perform across various workloads. But designing a model that can be simplified is extremely challenging. Recently, AutoML algorithms have emerged to help researchers find the right model without the need for manual experimentation. However, more often than not, these algorithms are compute-heavy and need thousands of models to train.

Model Search — which is built on Google’s TensorFlow ML framework. It can either run on a single machine or several, it consists of multiple trainers — a transfer learning algorithm, a search algorithm, and a database to store evaluated models. Model Search runs training and evaluation experiments for Artificial Intelligence models in an adaptive and asynchronous fashion, such that all trainers share the knowledge gained from their experiments while carrying on each experiment independently. At the beginning of every cycle, the search algorithm looks up all the completed trials and decides what to try next, after which it “mutates” over one of the best architectures found up to that point and assigns the resulting model back to a trainer.

To further improve efficiency and accuracy, Model Search employs transfer learning during experiments. For instance, it uses knowledge distillation and weight sharing, which bootstraps some of the variables in models from formerly trained models. This allows faster training and by extension opportunities to learn more and apparently better architectures.

After a Model Search run, users can compare the many models found during the search. Additionally, they can make their own search space to modify the architectural elements in their models.

Google says that in an internal experiment, Model Search improved upon production models with negligible iterations, mainly in the areas of keyword spotting and language identification. It also managed to find an architecture appropriate for image classification on the heavily explored CIFAR-10 open-source imaging dataset.

Google research engineer Hanna Mazzawi and research scientist Xavi Gonzalvo wrote in a blog post. “We hope the Model Search code will provide researchers with a flexible, domain-agnostic framework for machine learning model discovery. By building upon previous knowledge for a given domain, we believe that this framework is powerful enough to build models with state-of-the-art performance on well-studied problems when provided with a search space composed of standard building blocks.”

About Google

A problem isn't truly solved until it's solved for all. Googlers build products that help create opportunities for everyone, whether down the street or across the globe. Bring your insight, imagination and a healthy disregard for the impossible. Bring everything that makes you unique. Together, we can build for everyone.

Spotlight

Other News
AI Tech

AI and Big Data Expo North America announces leading Speaker Lineup

TechEx Events | March 07, 2024

AI and Big Data Expo North America announces new speakers! SANTA CLARA, CALIFORNIA, UNITED STATES, February 26, 2024 /EINPresswire.com/ -- TheAI and Big Expo North America, the leading event for Enterprise AI, Machine Learning, Security, Ethical AI, Deep Learning, Data Ecosystems, and NLP, has announced a fresh cohort of distinguishedspeakersfor its upcoming conference at the Santa Clara Convention Center on June 5-6, 2024. Some of the top industry speakers set to take the stage are: - Sam Hamilton - Head of Data & AI – Visa - Dr Astha Purohit - Director - Product (Tech) Ops – Walmart - Noorddin Taj - Head of Architecture and Design of Intelligent Operations - BP - Temi Odesanya - Director - AI Governance Automation - Thomson Reuters - Katie Sanders - Assistant Vice President – Tech - Union Pacific Railroad - Prasanth Nandanuru – SVP - Wells Fargo - Rodney Brooks - Professor Emeritus - MIT These esteemed speakers bring a wealth of knowledge and expertise to an already impressive lineup, promising attendees a truly enlightening experience. In addition to the speakers, theAI and Big Data Expo North Americawill feature a series of presentations covering a diverse range of topics in AI and Big Data exploring the latest innovations, implementations and strategies across a range of industries. Attendees can expect to gain valuable insights and practical strategies from presentations such as: How Gen AI Positively Augments Workforce Capabilities Trends in Computer Vision: Applications, Datasets, and Models Getting to Production-Ready: Challenges and Best Practices for Deploying AI Ensuring Your AI is Responsible and Ethical Mitigating Bias and Promoting Fairness in AI Systems Security Challenges in the Era of Gen AI and Data Science AI for Good: Social Impact and Ethics Selling Data Democratization to Executives Spreading Data Insights across the Business Barriers to Overcome: People, Processes, and Technology Optimizing the Customer Experience with AI Using AI to Drive Growth in a Regulated Industry Building an MLOps Foundation for AI at Scale The Expo offers a platform for exploration and discovery, showcasing how cutting-edge technologies are reshaping a myriad of industries, including manufacturing, transport, supply chain, government, legal sectors, financial services, energy, utilities, insurance, healthcare, retail, and more. Attendees will have the chance to witness firsthand the transformative power of AI and Big Data across various sectors, gaining insights that are crucial for staying ahead in today's rapidly evolving technological landscape. Anticipating a turnout of over 7000 attendees and featuring 200 speakers across various tracks, AI and Big Data Expo North America offers a unique opportunity for CTO’s, CDO’s, CIO’s , Heads of IOT, AI /ML, IT Directors and tech enthusiasts to stay abreast of the latest trends and innovations in AI, Big Data and related technologies. Organized by TechEx Events, the conference will also feature six co-located events, including the IoT Tech Expo, Intelligent Automation Conference, Cyber Security & Cloud Congress, Digital Transformation Week, and Edge Computing Expo, ensuring a comprehensive exploration of the technological landscape. Attendees can choose from various ticket options, providing access to engaging sessions, the bustling expo floor, premium tracks featuring industry leaders, a VIP networking party, and a sophisticated networking app facilitating connections ahead of the event. Secure your ticket with a 25% discount on tickets, available until March 31st, 2024. Save up to $300 on your ticket and be part of the conversation shaping the future of AI and Big Data technologies. For more information and to secure your place at AI and Big Data Expo North America, please visit https://www.ai-expo.net/northamerica/. About AI and Big Data Expo North America: The AI and Big Data Expo North America is a leading event in the AI and Big Data landscape, serving as a nexus for professionals, industry experts, and enthusiasts to explore and navigate the ever-evolving technological frontier. Through its focus on education, networking, and collaboration, the Expo continues to be a beacon for those eager to stay at the forefront of technological innovation. “AI and Big Data Expo North Americais a part ofTechEx. For more information regardingTechExplease see onlinehere.”

Read More