scorecardresearch
Follow Us:
Tuesday, May 18, 2021

IISc team studies differences between mechanized brain and human brain to equip them with similar traits

The senior author of the study added that such networks were still nowhere close to performing as well as the human brain in perceiving visual cues, despite having evolved significantly over the past decade.

By: Express Web Desk | Bengaluru |
April 21, 2021 4:17:10 pm
BrainThe current study has been published in Nature Communications, a widely recognised peer-reviewed journal that covers topics in natural sciences, including physics, chemistry, earth sciences, medicine, and biology. (Representational)

A team of researchers at the Indian Institute of Science (IISc) Bengaluru has discovered remarkable differences between deep neural networks — machine learning systems inspired by the network of brain cells or neurons in the human brain — and the human brain itself, which they believe would help in bridging the characteristic gaps between them.

According to SP Arun, associate professor at Centre for Neuroscience (CNS) at IISc, many studies showing similarities between deep networks and brains were already available. “However, none has really looked at their systematic differences, which would help us get closer to making these networks more brain-like,” he said.

Also the senior author of the study which compared the visual perception between neural networks and the human brain, Arun added that such networks were still nowhere close to performing as well as the human brain in perceiving visual cues, despite having evolved significantly over the past decade.

Explaining how the study was carried out, PhD student and first author of the study Georgin Jacob said, “Neural networks, unlike the brain, focussed on the finer details of an image first”. This, he said, helped the team understand that these networks and the human brain followed very different steps in carrying out the same object recognition tasks like identifying details of a face.

Jacob added that humans first look at the face as a whole and then focus on finer details like the eyes, nose, mouth and so on when presented with an image, while the same for neural networks took place in the inverse order.

Further, the team studied 13 different perceptual effects and uncovered previously unknown qualitative differences between deep networks and the human brain. “An example is the Thatcher effect, a phenomenon where humans find it easier to recognise local feature changes in an upright image, but this becomes difficult when the image is flipped upside-down. Deep networks trained to recognise upright faces showed a Thatcher effect when compared with networks trained to recognise objects,” an IISc release mentioned.

The current study has been published in Nature Communications, a widely recognised peer-reviewed journal that covers topics in natural sciences, including physics, chemistry, earth sciences, medicine, and biology.

“Such analyses can help researchers build more robust neural networks that not only perform better but are also immune to ‘adversarial attacks’ that aim to derail them,” the Institute added.

📣 The Indian Express is now on Telegram. Click here to join our channel (@indianexpress) and stay updated with the latest headlines

For all the latest Bangalore News, download Indian Express App.

  • The Indian Express website has been rated GREEN for its credibility and trustworthiness by Newsguard, a global service that rates news sources for their journalistic standards.
Advertisement
Advertisement
Advertisement
Advertisement
X
x