{"id":18384,"date":"2025-02-01T00:12:17","date_gmt":"2025-02-01T00:12:17","guid":{"rendered":"https:\/\/enitajobs.com\/employer\/ptrevival\/"},"modified":"2025-02-01T00:34:58","modified_gmt":"2025-02-01T00:34:58","slug":"walkandtalkrentals","status":"publish","type":"employer","link":"https:\/\/enitajobs.com\/en\/employer\/walkandtalkrentals\/","title":{"rendered":"Walkandtalkrentals"},"content":{"rendered":"

MIT Researchers Develop an Efficient Way to Train more Reliable AI Agents<\/strong><\/p>\n

<\/p>\n

Fields ranging from robotics to medication to political science are attempting to train AI<\/a> systems to make significant decisions<\/a> of all kinds. For instance, utilizing an AI<\/a> system to wisely control traffic in a congested city<\/a> could help motorists reach their locations quicker, while improving security or sustainability.<\/p>\n

<\/p>\n

Unfortunately, teaching an AI<\/a> system to make great decisions is no simple job.<\/p>\n

Reinforcement learning models, which underlie these AI<\/a> decision-making systems, still often fail when faced with even small variations in the tasks they are trained to perform. In the case of traffic, a design might struggle to manage a set of crossways with different speed limitations, varieties of lanes, or traffic patterns<\/a>.<\/p>\n

To increase the reliability of support learning designs for intricate tasks with variability, MIT researchers have actually introduced a more efficient algorithm for training them.<\/p>\n

The algorithm tactically chooses the best jobs for training an AI<\/a> agent so it can successfully carry out all tasks in a collection of related tasks. In the case of traffic signal control, each job might be one crossway in a task area that consists of all crossways in the city.<\/p>\n

By focusing on a smaller sized number of intersections that contribute the most to the algorithm’s overall efficiency, this method optimizes performance while keeping the training cost low.<\/p>\n

The scientists discovered that their method was between five and 50 times more effective than basic techniques on a selection of simulated tasks. This gain in effectiveness assists the algorithm learn a much better option in a much faster way, eventually improving the efficiency of the AI<\/a> agent.<\/p>\n

“We were able to see extraordinary performance improvements, with a really simple algorithm, by believing outside the box. An algorithm that is not very complex stands a better chance of being embraced by the neighborhood because it is simpler to implement and much easier for others to understand,” says senior author Cathy Wu, the Thomas D. and Virginia W. Cabot Career Development Associate Professor in Civil and Environmental Engineering (CEE) and the Institute for Data, Systems, and Society (IDSS), and a member of the Laboratory for Information and Decision Systems (LIDS).<\/p>\n

She is signed up with on the paper by lead author Jung-Hoon Cho, a CEE college student; Vindula Jayawardana, a college student in the Department of Electrical Engineering and Computer Technology (EECS); and Sirui Li, an IDSS college . The research study will exist at the Conference on Neural Information Processing Systems.<\/p>\n

Finding a middle ground<\/p>\n

To train an algorithm to control traffic signal at numerous crossways in a city, an engineer would usually choose between two primary techniques<\/a>. She can train one algorithm for each intersection individually, utilizing only that intersection’s data, or train a larger algorithm using data from all crossways and after that apply it to each one.<\/p>\n

But each method includes its share of disadvantages. Training a separate algorithm for each job (such as a given intersection) is a time-consuming process that requires a huge quantity of information and calculation, while training one algorithm for all tasks typically results in subpar efficiency.<\/p>\n

Wu and her collaborators looked for a sweet spot in between these two techniques.<\/p>\n

<\/p>\n

For their technique, they select a subset of tasks and train one algorithm for each task independently. Importantly, they tactically select individual tasks which are more than likely to enhance the algorithm’s overall performance on all tasks.<\/p>\n

They utilize a typical technique from the reinforcement learning field called zero-shot transfer knowing, in which a currently trained model is applied to a new job without being additional trained<\/a>. With transfer knowing, the model frequently carries out remarkably well on the brand-new next-door neighbor<\/a> task.<\/p>\n

“We understand it would be perfect to train on all the jobs, but we wondered if we could get away with training on a subset of those jobs, apply the result to all the tasks, and still see a performance boost,” Wu states.<\/p>\n

To determine which jobs they ought to choose to take full advantage of anticipated performance, the researchers developed an algorithm called Model-Based Transfer Learning (MBTL).<\/p>\n

The MBTL algorithm has two pieces. For one, it models how well each algorithm would carry out if it were trained independently on one job. Then it models how much each algorithm’s performance would break down if it were moved to each other job, an idea referred to as generalization performance.<\/p>\n

Explicitly modeling generalization efficiency permits MBTL to approximate the worth of training on a new task.<\/p>\n

MBTL does this sequentially, choosing the task which causes the highest efficiency gain first, then selecting<\/a> additional tasks that offer the most significant subsequent marginal<\/a> improvements to overall performance.<\/p>\n

Since MBTL just focuses on the most promising jobs, it can significantly enhance the effectiveness of the training procedure.<\/p>\n

Reducing training expenses<\/p>\n

When the researchers evaluated this method on simulated tasks, including managing traffic signals, handling real-time speed advisories, and carrying out several traditional control tasks, it was 5 to 50 times more effective than other approaches.<\/p>\n

This suggests they could show up at the very same solution<\/a> by training on far less data. For example, with a 50x effectiveness increase, the MBTL algorithm might train on just 2 tasks and accomplish the very same performance as a standard method which uses data from 100 tasks.<\/p>\n

“From the point of view of the two main methods, that implies data from the other 98 jobs was not required or that training on all 100 jobs is puzzling to the algorithm, so the performance winds up worse than ours,” Wu says.<\/p>\n

With MBTL, adding even a percentage of extra training time could cause far better performance.<\/p>\n

In the future, the researchers prepare to create MBTL algorithms that can reach more complicated issues, such as high-dimensional job spaces. They are likewise thinking about using their approach to real-world problems, especially in next-generation mobility<\/a> systems.<\/p>\n

<\/p>\n","protected":false},"featured_media":0,"comment_status":"open","ping_status":"closed","template":"","employer_category":[],"employer_location":[],"class_list":["post-18384","employer","type-employer","status-publish","hentry"],"cmb2":{"_employer_general":{"_employer_attached_user":"","_employer_email":"","_employer_founded_date":"","_employer_website":"","_employer_phone":"","_employer_featured":"","_employer_cover_photo":"","_employer_cover_photo_id":"","_employer_profile_photos":"","_employer_video_url":"","_employer_layout_type":""},"_employer_socials":{"_employer_socials":""},"_employer_map_location":{"_employer_address":"","_employer_map_location":""},"_employer_team_members":{"_employer_team_members":""},"_employer_employees":{"_employer_employees":[]}},"_links":{"self":[{"href":"https:\/\/enitajobs.com\/en\/wp-json\/wp\/v2\/employer\/18384","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/enitajobs.com\/en\/wp-json\/wp\/v2\/employer"}],"about":[{"href":"https:\/\/enitajobs.com\/en\/wp-json\/wp\/v2\/types\/employer"}],"replies":[{"embeddable":true,"href":"https:\/\/enitajobs.com\/en\/wp-json\/wp\/v2\/comments?post=18384"}],"wp:attachment":[{"href":"https:\/\/enitajobs.com\/en\/wp-json\/wp\/v2\/media?parent=18384"}],"wp:term":[{"taxonomy":"employer_category","embeddable":true,"href":"https:\/\/enitajobs.com\/en\/wp-json\/wp\/v2\/employer_category?post=18384"},{"taxonomy":"employer_location","embeddable":true,"href":"https:\/\/enitajobs.com\/en\/wp-json\/wp\/v2\/employer_location?post=18384"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}