Home » A liquid neural network that learns on the job, not just during its training phase
Ongoing Explorations

A liquid neural network that learns on the job, not just during its training phase

MIT researchers have developed neural “liquid” network that varies its equations parameters, enhancing its ability to analyze time series data.

Image: Jose-Luis Olivares, MIT


Exploration Highlights


  • The new type of neural network could aid decision making in autonomous driving and medical diagnosis.
  • These flexible algorithms, dubbed “liquid” 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.
  • “This is a way forward for the future of robot control, natural language processing, video processing — any form of time series data processing,” says Ramin Hasani
  • Hasani designed a neural network that can adapt to the variability of real-world systems.
  • This flexibility is key. Most neural networks’ behavior is fixed after the training phase, which means they’re bad at adjusting to changes in the incoming data stream. Hasani says the fluidity of his “liquid” 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. “So, it’s more robust,” he says.
  • Hasani’s 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.
  • “We think this kind of network could be a key element of future intelligence systems.”


Original Article: “Liquid” machine-learning system adapts to changing conditions

More from: Massachusetts Institute of Technology on Innovation2


Go deeper with the LATEST Bing News on:

Liquid neural network
  • John Deere Debuts Autonomous Tractor: “One Giant Robot”
    on January 20, 2022 at 10:28 pm

    and a deep neural network that classifies each pixel of each image in 100 milliseconds. This determines if the machine continues to move or stops when an obstacle is detected. Of course, John Deere is ...

  • AKVIS Noise Buster AI 12.0: AI-Powered Noise Reduction Technologies to Enhance Photos!
    on January 20, 2022 at 5:45 am

    AKVIS is excited to announce the release of AKVIS Noise Buster AI 12.0, the AI-powered noise reduction software. Version 12.0 provides trained neural networks, a redesigned user interface, support for ...

  • PsyLink An Open Source Neural Interface For Non-Invasive EMG
    on January 19, 2022 at 3:59 pm

    We don’t see many EMG (electromyography) projects, despite how cool the applications can be. This may be because of technical difficulties with seeing the tiny muscular electrical signals ...

  • How well do explanation methods for machine-learning models work?
    on January 18, 2022 at 7:17 am

    Imagine a team of physicians using a neural network to detect cancer in mammogram images. Even if this machine-learning model seems to be performing well, it might be focusing on image features that ...

  • The best iPhone deals you can get in January 2022
    on January 13, 2022 at 4:01 pm

    In general, we find that you'll get a better deal away from official network stores (although it ... Features: 6.1” Liquid Retina HD display1, Glass and aluminium design, Dual 12MP Ultra Wide ...

  • Tesla's cameras-only autonomous system stirs controversy
    on January 6, 2022 at 1:27 pm

    Tesla Chief Executive Elon Musk has touted a system built around eight "surround" cameras that feed data into the auto's "deep neural network," according to Tesla's website. But as with so many other ...

  • neural network
    on January 4, 2022 at 2:07 am

    Berlin-based Jina.ai, an open source startup that uses neural search to help its users find information in their unstructured data (including videos and images), today announced that it has raised a $ ...

  • 10 Best Laptops for Video Editing in India (January 2022) - Buyer’s Guide
    on January 2, 2022 at 3:59 pm

    And accelerate machine learning tasks with the 16-core Neural Engine. All with a silent, fanless design and the longest battery life ever up to 18 hours. The HP Spectre series is known as the most ...

  • 10 of the weirdest experiments of 2021
    on December 30, 2021 at 6:09 am

    They made the Cronenberg-esque mini brain, called an organoid, by transforming stem cells into neural tissue, then stimulating the cells with chemical signals to form tiny rudimentary "optic cups ...

  • Electrical and Computer Engineering
    on December 11, 2021 at 5:25 pm

    Topics include truth and underdetermination, induction, computability, language learning, pattern recognition, neural networks, and the role of simplicity ... junction lasers, liquid crystal devices, ...

Go deeper with the LATEST Bing News on:

Neural network




Add Comment

Click here to post a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.