New AI algorithm taught by humans learns beyond its training

November 16, 2016
This figure compares a traditionally trained algorithm to Aarabi and Guo's heuristically trained neural net. The left and centre columns show an aggressive and conservative image-recognition algorithm trained to recognized human hair, compared to the more precise heuristically trained algorithm at right. Credit: Courtesy: IEEE Trans NN & LS

"Hey Siri, how's my hair?" Your smartphone may soon be able to give you an honest answer, thanks to a new machine learning algorithm designed by U of T Engineering researchers Parham Aarabi and Wenzhi Guo.

The team designed an that learns directly from human instructions, rather than an existing set of examples, and outperformed conventional methods of training neural networks by 160 per cent. But more surprisingly, their algorithm also outperformed its own training by nine per cent—it learned to recognize hair in pictures with greater reliability than that enabled by the training, marking a significant leap forward for .

Aarabi and Guo trained their algorithm to identify people's hair in photographs—a much more challenging task for computers than it is for humans.

"Our algorithm learned to correctly classify difficult, borderline cases—distinguishing the texture of hair versus the texture of the background," says Aarabi. "What we saw was like a teacher instructing a child, and the child learning beyond what the teacher taught her initially."

Humans "teach" neural networks—computer networks that learn dynamically—by providing a set of labeled data and asking the neural network to make decisions based on the samples it's seen. For example, you could train a neural network to identify sky in a photograph by showing it hundreds of pictures with the sky labeled.

This algorithm is different: it learns directly from human trainers. With this model, called heuristic training, humans provide direct instructions that are used to pre-classify training samples rather than a set of fixed examples. Trainers program the algorithm with guidelines such as "Sky is likely to be varying shades of blue," and "Pixels near the top of the image are more likely to be sky than pixels at the bottom."

Their work is published in the journal IEEE Transactions on Neural Networks and Learning Systems.

This heuristic training approach holds considerable promise for addressing one of the biggest challenges for : making correct classifications of previously unknown or unlabeled data. This is crucial for applying machine learning to new situations, such as correctly identifying cancerous tissues for medical diagnostics, or classifying all the objects surrounding and approaching a self-driving car.

"Applying heuristic training to hair segmentation is just a start," says Guo. "We're keen to apply our method to other fields and a range of applications, from medicine to transportation."

Explore further: Students develop algorithm that categorizes user-generated restaurant photos

More information: Wenzhangzhi Guo et al, Hair Segmentation Using Heuristically-Trained Neural Networks, IEEE Transactions on Neural Networks and Learning Systems (2016). DOI: 10.1109/TNNLS.2016.2614653

Related Stories

Engineers teach machines to recognize tree species

September 14, 2016

Engineers from Caltech have developed a method that uses data from satellite and street-level images, such as the ones that you can see in Google maps, to create automatically an inventory of street trees that cities may ...

Recommended for you

Firms push hydrogen as top green energy source

January 18, 2017

Over a dozen leading European and Asian firms have teamed up to promote the use of hydrogen as a clean fuel and cut the production of harmful gasses that lead to global warming.

WhatsApp vulnerable to snooping: report

January 13, 2017

The Facebook-owned mobile messaging service WhatsApp is vulnerable to interception, the Guardian newspaper reported on Friday, sparking concern over an app advertised as putting an emphasis on privacy.

2 comments

Adjust slider to filter visible comments by rank

Display comments: newest first

Andrew Palfreyman
1 / 5 (2) Nov 17, 2016
Jesus. What is this? - the B Ark? How about designing better jet engines or other useful pursuits?
EyeNStein
2 / 5 (1) Nov 19, 2016
Its obvious looking at their pictures of hair that the AI has no concept of hair being a fine stranded thing. Its colouring the wavy blobby textures that are emergent. (And making a hash of that task too.)

This looks like a basic undergraduate project on pattern recognition. Its about as effective as an AI that says "its favourite colour is blue", without knowing what a colour is, incapable of having any "favourite" or actually recognising blue.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.