{"id":19758,"date":"2021-01-29T08:20:00","date_gmt":"2021-01-29T02:35:00","guid":{"rendered":"https:\/\/www.revoscience.com\/en\/?p=19758"},"modified":"2021-01-28T23:33:43","modified_gmt":"2021-01-28T17:48:43","slug":"liquid-machine-learning-system-adapts-to-changing-conditions","status":"publish","type":"post","link":"https:\/\/www.revoscience.com\/en\/liquid-machine-learning-system-adapts-to-changing-conditions\/","title":{"rendered":"\u201cLiquid\u201d machine-learning system adapts to changing conditions"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

CAMBRIDGE, Mass. (MIT News Office) —\u00a0MIT researchers have developed a type of neural network that learns on the job, not just during its training phase. These flexible algorithms, dubbed \u201cliquid\u201d networks, change their underlying equations to continuously adapt to new data inputs. The advance could aid decision making based on data streams that change over time, including those involved in medical diagnosis and autonomous driving.<\/p>\n\n\n\n

The new type of neural network could aid decision making in autonomous driving and medical diagnosis.<\/strong><\/em><\/p><\/blockquote>\n\n\n\n

\u201cThis is a way forward for the future of robot control, natural language processing, video processing \u2014 any form of time series data processing,\u201d says Ramin Hasani, the study\u2019s lead author. \u201cThe potential is really significant.\u201d<\/p>\n\n\n\n

The research will be presented at February\u2019s AAAI Conference on Artificial Intelligence. In addition to Hasani, a postdoc in the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), MIT co-authors include Daniela Rus, CSAIL director and the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science, and PhD student Alexander Amini. Other co-authors include Mathias Lechner of the Institute of Science and Technology Austria and Radu Grosu of the Vienna University of Technology.<\/p>\n\n\n\n

Time series data are both ubiquitous and vital to our understanding the world, according to Hasani. \u201cThe real world is all about sequences. Even our perception \u2014 you\u2019re not perceiving images, you\u2019re perceiving sequences of images,\u201d he says. \u201cSo, time series data actually create our reality.\u201d<\/p>\n\n\n\n

He points to video processing, financial data, and medical diagnostic applications as examples of time series that are central to society. The vicissitudes of these ever-changing data streams can be unpredictable. Yet analyzing these data in real time, and using them to anticipate future behavior, can boost the development of emerging technologies like self-driving cars. So Hasani built an algorithm fit for the task.<\/p>\n\n\n\n

Hasani designed a neural network that can adapt to the variability of real-world systems. Neural networks are algorithms that recognize patterns by analyzing a set of \u201ctraining\u201d examples. They\u2019re often said to mimic the processing pathways of the brain \u2014 Hasani drew inspiration directly from the microscopic nematode, C. elegans<\/em>. \u201cIt only has 302 neurons in its nervous system,\u201d he says, \u201cyet it can generate unexpectedly complex dynamics.\u201d<\/p>\n\n\n\n

Hasani coded his neural network with careful attention to how C. elegans<\/em> neurons activate and communicate with each other via electrical impulses. In the equations he used to structure his neural network, he allowed the parameters to change over time based on the results of a nested set of differential equations.<\/p>\n\n\n\n

This flexibility is key. Most neural networks\u2019 behavior is fixed after the training phase, which means they\u2019re bad at adjusting to changes in the incoming data stream. Hasani says the fluidity of his \u201cliquid\u201d network makes it more resilient to unexpected or noisy data, like if heavy rain obscures the view of a camera on a self-driving car. \u201cSo, it\u2019s more robust,\u201d he says.<\/p>\n\n\n\n

There\u2019s another advantage of the network\u2019s flexibility, he adds: \u201cIt\u2019s more interpretable.\u201d<\/p>\n\n\n\n

Hasani says his liquid network skirts the inscrutability<\/a> common to other neural networks. \u201cJust changing the representation of a neuron,\u201d which Hasani did with the differential equations, \u201cyou can really explore some degrees of complexity you couldn\u2019t explore otherwise.\u201d Thanks to Hasani\u2019s small number of highly expressive neurons, it\u2019s easier to peer into the \u201cblack box\u201d of the network\u2019s decision making and diagnose why the network made a certain characterization.<\/p>\n\n\n\n

\u201cThe model itself is richer in terms of expressivity,\u201d says Hasani. That could help engineers understand and improve the liquid network\u2019s performance.<\/p>\n\n\n\n

Hasani\u2019s network excelled in a battery of tests. It edged out other state-of-the-art time series algorithms by a few percentage points in accurately predicting future values in datasets, ranging from atmospheric chemistry to traffic patterns. \u201cIn many applications, we see the performance is reliably high,\u201d he says. Plus, the network\u2019s small size meant it completed the tests without a steep computing cost. \u201cEveryone talks about scaling up their network,\u201d says Hasani. \u201cWe want to scale down, to have fewer but richer nodes.\u201d<\/p>\n\n\n\n

Hasani plans to keep improving the system and ready it for industrial application. \u201cWe have a provably more expressive neural network that is inspired by nature. But this is just the beginning of the process,\u201d he says. \u201cThe obvious question is how do you extend this? We think this kind of network could be a key element of future intelligence systems.\u201d<\/p>\n\n\n\n

This research was funded, in part, by Boeing, the National Science Foundation, the Austrian Science Fund, and Electronic Components and Systems for European Leadership.<\/p>\n","protected":false},"excerpt":{"rendered":"

MIT researchers have developed a type of neural network that learns on the job, not just during its training phase. These flexible algorithms, dubbed \u201cliquid\u201d networks, change their underlying equations to continuously adapt to new data inputs. <\/p>\n","protected":false},"author":2,"featured_media":19759,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[47,17],"tags":[],"class_list":["post-19758","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-it","category-research"],"featured_image_urls":{"full":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",1000,667,false],"thumbnail":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press-200x200.jpg",200,200,true],"medium":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press-600x400.jpg",600,400,true],"medium_large":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press-768x512.jpg",750,500,true],"large":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press-675x450.jpg",675,450,true],"1536x1536":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",1000,667,false],"2048x2048":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",1000,667,false],"ultp_layout_landscape_large":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",1000,667,false],"ultp_layout_landscape":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",855,570,false],"ultp_layout_portrait":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",600,400,false],"ultp_layout_square":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",600,400,false],"newspaper-x-single-post":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press-760x490.jpg",760,490,true],"newspaper-x-recent-post-big":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press-550x360.jpg",550,360,true],"newspaper-x-recent-post-list-image":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press-95x65.jpg",95,65,true],"web-stories-poster-portrait":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",640,427,false],"web-stories-publisher-logo":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",96,64,false],"web-stories-thumbnail":["https:\/\/www.revoscience.com\/en\/wp-content\/uploads\/2021\/01\/MIT-liquid-networks-press.jpg",150,100,false]},"author_info":{"info":["RevoScience"]},"category_info":"IT<\/a> Research<\/a>","tag_info":"Research","comment_count":"0","_links":{"self":[{"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/posts\/19758","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/comments?post=19758"}],"version-history":[{"count":0,"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/posts\/19758\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/media\/19759"}],"wp:attachment":[{"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/media?parent=19758"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/categories?post=19758"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.revoscience.com\/en\/wp-json\/wp\/v2\/tags?post=19758"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}