̽��ֱ�� of Cambridge - deep learning

AI trained to identify least green homes by Cambridge researchers

ta385 — Thu, 02 Nov 2023 08:00:00 +0000

‘Hard-to-decarbonize’ (HtD) houses are responsible for over a quarter of all direct housing emissions – a major obstacle to achieving net zero – but are rarely identified or targeted for improvement.

Now a new ‘deep learning’ model trained by researchers from Cambridge ̽��ֱ��’s Department of Architecture promises to make it far easier, faster and cheaper to identify these high priority problem properties and develop strategies to improve their green credentials.

Houses can be ‘hard to decarbonize’ for various reasons including their age, structure, location, social-economic barriers and availability of data. Policymakers have tended to focus mostly on generic buildings or specific hard-to-decarbonise technologies but the study, published in the journal Sustainable Cities and Society, could help to change this.

Maoran Sun, an urban researcher and data scientist, and his PhD supervisor Dr Ronita Bardhan (Selwyn College), who leads Cambridge’s Sustainable Design Group, show that their AI model can classify HtD houses with 90% precision and expect this to rise as they add more data, work which is already underway.

Dr Bardhan said: “This is the first time that AI has been trained to identify hard-to-decarbonize buildings using open source data to achieve this.

“Policymakers need to know how many houses they have to decarbonize, but they often lack the resources to perform detail audits on every house. Our model can direct them to high priority houses, saving them precious time and resources.”

̽��ֱ��model also helps authorities to understand the geographical distribution of HtD houses, enabling them to efficiently target and deploy interventions efficiently.

̽��ֱ��researchers trained their AI model using data for their home city of Cambridge, in the United Kingdom. They fed in data from Energy Performance Certificates (EPCs) as well as data from street view images, aerial view images, land surface temperature and building stock. In total, their model identified 700 HtD houses and 635 non-HtD houses. All of the data used was open source.

Maoran Sun said: “We trained our model using the limited EPC data which was available. Now the model can predict for the city’s other houses without the need for any EPC data.”

Bardhan added: “This data is available freely and our model can even be used in countries where datasets are very patchy. ̽��ֱ��framework enables users to feed in multi-source datasets for identification of HtD houses.”

Sun and Bardhan are now working on an even more advanced framework which will bring additional data layers relating to factors including energy use, poverty levels and thermal images of building facades. They expect this to increase the model’s accuracy but also to provide even more detailed information.

̽��ֱ��model is already capable of identifying specific parts of buildings, such as roofs and windows, which are losing most heat, and whether a building is old or modern. But the researchers are confident they can significantly increase detail and accuracy.

They are already training AI models based on other UK cities using thermal images of buildings, and are collaborating with a space products-based organisation to benefit from higher resolution thermal images from new satellites. Bardhan has been part of the NSIP – UK Space Agency program where she collaborated with the Department of Astronomy and Cambridge Zero on using high resolution thermal infrared space telescopes for globally monitoring the energy efficiency of buildings.

Sun said: “Our models will increasingly help residents and authorities to target retrofitting interventions to particular building features like walls, windows and other elements.”

Bardhan explains that, until now, decarbonization policy decisions have been based on evidence derived from limited datasets, but is optimistic about AI’s power to change this.

“We can now deal with far larger datasets. Moving forward with climate change, we need adaptation strategies based on evidence of the kind provided by our model. Even very simple street view photographs can offer a wealth of information without putting anyone at risk.”

̽��ֱ��researchers argue that by making data more visible and accessible to the public, it will become much easier to build consensus around efforts to achieve net zero.

“Empowering people with their own data makes it much easier for them to negotiate for support,” Bardhan said.

She added: “There is a lot of talk about the need for specialised skills to achieve decarbonisation but these are simple data sets and we can make this model very user friendly and accessible for the authorities and individual residents.”

Cambridge as a study site

Cambridge is an atypical city but informative site on which to base the initial model. Bardhan notes that Cambridge is relatively affluent meaning that there is a greater willingness and financial ability to decarbonise houses.

“Cambridge isn’t ‘hard to reach’ for decarbonisation in that sense,” Bardhan said. “But the city’s housing stock is quite old and building bylaws prevent retrofitting and the use of modern materials in some of the more historically important properties. So it faces interesting challenges.”

̽��ֱ��researchers will discuss their findings with Cambridge City Council. Bardhan previously worked with the Council to assess council houses for heat loss. They will also continue to work with colleagues at Cambridge Zero and the ̽��ֱ��’s Decarbonisation Network.

Reference

M Sun & R Bardhan, ‘Identifying Hard-to-Decarbonize houses from multi-source data in Cambridge, UK’, Sustainable Cities and Society (2023). DOI: 10.1016/j.scs.2023.105015

First of its kind AI-model can help policymakers efficiently identify and prioritize houses for retrofitting and other decarbonizing measures.

This is the first time that AI has been trained to identify hard-to-decarbonize buildings

Ronita Bardhan

Street view images of houses in Cambridge, UK, identifying building features. Red represents region contributing most to the 'Hard-to-decarbonize' identification. Blue represents low contribution.

̽��ֱ��text in this work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Images, including our videos, are Copyright © ̽��ֱ�� of Cambridge and licensors/contributors as identified. All rights reserved. We make our image and video content available in a number of ways – as here, on our main website under its Terms and conditions, and on a range of channels including social media that permit your use and sharing of our content under their respective Terms.

Yes

Licence type:

Attribution-Noncommerical

How sure is sure? Incorporating human error into machine learning

sc604 — Wed, 09 Aug 2023 23:40:58 +0000

Human error and uncertainty are concepts that many artificial intelligence systems fail to grasp, particularly in systems where a human provides feedback to a machine learning model. Many of these systems are programmed to assume that humans are always certain and correct, but real-world decision-making includes occasional mistakes and uncertainty.

Researchers from the ̽��ֱ�� of Cambridge, along with ̽��ֱ��Alan Turing Institute, Princeton, and Google DeepMind, have been attempting to bridge the gap between human behaviour and machine learning, so that uncertainty can be more fully accounted for in AI applications where humans and machines are working together. This could help reduce risk and improve trust and reliability of these applications, especially where safety is critical, such as medical diagnosis.

̽��ֱ��team adapted a well-known image classification dataset so that humans could provide feedback and indicate their level of uncertainty when labelling a particular image. ̽��ֱ��researchers found that training with uncertain labels can improve these systems’ performance in handling uncertain feedback, although humans also cause the overall performance of these hybrid systems to drop. Their results will be reported at the AAAI/ACM Conference on Artificial Intelligence, Ethics and Society (AIES 2023) in Montréal.

‘Human-in-the-loop’ machine learning systems – a type of AI system that enables human feedback – are often framed as a promising way to reduce risks in settings where automated models cannot be relied upon to make decisions alone. But what if the humans are unsure?

“Uncertainty is central in how humans reason about the world but many AI models fail to take this into account,” said first author Katherine Collins from Cambridge’s Department of Engineering. “A lot of developers are working to address model uncertainty, but less work has been done on addressing uncertainty from the person’s point of view.”

We are constantly making decisions based on the balance of probabilities, often without really thinking about it. Most of the time – for example, if we wave at someone who looks just like a friend but turns out to be a total stranger – there’s no harm if we get things wrong. However, in certain applications, uncertainty comes with real safety risks.

“Many human-AI systems assume that humans are always certain of their decisions, which isn’t how humans work – we all make mistakes,” said Collins. “We wanted to look at what happens when people express uncertainty, which is especially important in safety-critical settings, like a clinician working with a medical AI system.”

“We need better tools to recalibrate these models, so that the people working with them are empowered to say when they’re uncertain,” said co-author Matthew Barker, who recently completed his MEng degree at Gonville & Caius College, Cambridge. “Although machines can be trained with complete confidence, humans often can’t provide this, and machine learning models struggle with that uncertainty.”

For their study, the researchers used some of the benchmark machine learning datasets: one was for digit classification, another for classifying chest X-rays, and one for classifying images of birds. For the first two datasets, the researchers simulated uncertainty, but for the bird dataset, they had human participants indicate how certain they were of the images they were looking at: whether a bird was red or orange, for example. These annotated ‘soft labels’ provided by the human participants allowed the researchers to determine how the final output was changed. However, they found that performance degraded rapidly when machines were replaced with humans.

“We know from decades of behavioural research that humans are almost never 100% certain, but it’s a challenge to incorporate this into machine learning,” said Barker. “We’re trying to bridge the two fields so that machine learning can start to deal with human uncertainty where humans are part of the system.”

̽��ֱ��researchers say their results have identified several open challenges when incorporating humans into machine learning models. They are releasing their datasets so that further research can be carried out and uncertainty might be built into machine learning systems.

“As some of our colleagues so brilliantly put it, uncertainty is a form of transparency, and that’s hugely important,” said Collins. “We need to figure out when we can trust a model and when to trust a human and why. In certain applications, we’re looking at probability over possibilities. Especially with the rise of chatbots, for example, we need models that better incorporate the language of possibility, which may lead to a more natural, safe experience.”

“In some ways, this work raised more questions than it answered,” said Barker. “But even though humans may be miscalibrated in their uncertainty, we can improve the trustworthiness and reliability of these human-in-the-loop systems by accounting for human behaviour.”

̽��ֱ��research was supported in part by the Cambridge Trust, the Marshall Commission, the Leverhulme Trust, the Gates Cambridge Trust and the Engineering and Physical Sciences Research Council (EPSRC), part of UK Research and Innovation (UKRI).

Reference:
Katherine M Collins et al. ‘Human Uncertainty in Concept-Based AI Systems.’ Paper presented at the Sixth AAAI/ACM Conference on Artificial Intelligence, Ethics and Society (AIES 2023), August 8-10, 2023. Montréal, QC, Canada.

Researchers are developing a way to incorporate one of the most human of characteristics – uncertainty – into machine learning systems.

Uncertainty is central in how humans reason about the world but many AI models fail to take this into account

Katherine Collins

PeopleImages via Getty Images

Futuristic image of a doctor looking at brain scans

Yes

Women in STEM: Agnieszka Słowik

sc604 — Thu, 16 Jan 2020 07:00:00 +0000

Broadly, my research explores the reasoning capacity of neural networks. You might have seen these algorithms in action when using automatic face recognition on social media or issuing voice commands to your phone. Neural networks, also hidden behind media-friendly terms such as deep learning, are nowadays a go-to research direction if one is interested in attaining the state-of-the-art accuracy on a classification task associated with a large amount of data.

Despite their impressive practical performance, these models are limited in their ability to combine familiar ideas to arrive at new conclusions as they tend to simply memorise the data. Having learned from the examples of red squares and blue circles, a truly intelligent system surely shouldn’t be confused by a red circle. This is a core challenge in learning algorithms and I hope my research will contribute to the international efforts of the machine learning community to induce reasoning and generalisation in neural networks.

During my current internship at Mila Quebec AI Institute, I'm investigating how agents based on neural networks communicate with each other in order to solve simple games. These games draw inspiration from the studies on language evolution in humans. ̽��ֱ��communication aspect is particularly cool and exciting because by analysing the messages sent between the agents I can see how closely these algorithms mimic the reasoning process of a biological intelligent system.

I have been extremely fortunate with my supervisors (Mateja Jamnik and Sean Holden) as well as the welcoming and friendly nature of the Department of Computer Science and Technology. Cambridge provides students with a unique degree of freedom, independence and intellectual stimulation. What I particularly appreciate after my experience with competitive institutions in Poland and France is that Cambridge provides the best resources for obtaining a well-rounded education alongside the ‘hard skills’ in a student’s chosen field.

I’ve always liked the quote “the areas in which you struggle the most are the ones in which you have the most to give.” If you put a lot of effort into grasping a subject or solving a task that seems daunting to begin with, you are well-equipped to support others who struggle with the same task. I believe this also applies to challenges outside of research.

Embrace stepping out of the ‘good student’ role. ̽��ֱ��skills required in a research career, especially in science and technology, frequently won’t fully overlap with what led you to have the top grades in your previous education. Firstly, there won’t be nearly as much of the immediate positive feedback so it is crucial to enjoy the process apart from the results. Secondly, the work will never seem finished so it is more important to follow a healthy routine. Reach out to friendly experienced colleagues to find out how they cope with these challenges.

Work with a light and kind attitude to yourself and others. ̽��ֱ��trap of oscillating between imposter syndrome and ‘I’m like, a genius’ is real in research. At the end of the day you are learning, trying new things and having lots of fun, together with like-minded people.

Agnieszka Słowik is a PhD candidate in the Department of Computer Science and Technology, where she is a member of the artificial intelligence research group. Here, she tells us about neural networks and how they communicate with each other, the importance of supportive supervisors, and how to be a supportive team member.

̽��ֱ��text in this work is licensed under a Creative Commons Attribution 4.0 International License. Images, including our videos, are Copyright © ̽��ֱ�� of Cambridge and licensors/contributors as identified. All rights reserved. We make our image and video content available in a number of ways – as here, on our main website under its Terms and conditions, and on a range of channels including social media that permit your use and sharing of our content under their respective Terms.

Yes

AI crossword-solving application could make machines better at understanding language

sc604 — Mon, 07 Mar 2016 09:38:00 +0000

Researchers have designed a web-based platform which uses artificial neural networks to answer standard crossword clues better than existing commercial products specifically designed for the task. ̽��ֱ��system could help machines understand language more effectively.

In tests against commercial crossword-solving software, the system, designed by researchers from the UK, US and Canada, was more accurate at answering clues that were single words (e.g. ‘culpability’ – guilt), a short combination of words (e.g. ‘devil devotee’ – Satanist), or a longer sentence or phrase (e.g. ‘French poet and key figure in the development of Symbolism’ – Baudelaire). ̽��ֱ��system can also be used a ‘reverse dictionary’ in which the user describes a concept and the system returns possible words to describe that concept.

̽��ֱ��researchers used the definitions contained in six dictionaries, plus Wikipedia, to ‘train’ the system so that it could understand words, phrases and sentences – using the definitions as a bridge between words and sentences. Their results, published in the journal Transactions of the Association for Computational Linguistics, suggest that a similar approach may lead to improved output from more general language understanding and dialogue systems and information retrieval engines in general. All of the code and data behind the application has been made freely available for future research.

“Over the past few years, there’s been a mini-revolution in machine learning,” said Felix Hill of the ̽��ֱ�� of Cambridge’s Computer Laboratory, one of the paper’s authors. “We’re seeing a lot more usage of deep learning, which is especially useful for language perception and speech recognition.”

Deep learning refers to an approach in which artificial neural networks with little or no prior ‘knowledge’ are trained to recreate human abilities using massive amounts of data. For this particular application, the researchers used dictionaries – training the model on hundreds of thousands of definitions of English words, plus Wikipedia.

“Dictionaries contain just about enough examples to make deep learning viable, but we noticed that the models get better and better the more examples you give them,” said Hill. “Our experiments show that definitions contain a valuable signal for helping models to interpret and represent the meaning of phrases and sentences.”

Working with Anna Korhonen from the Cambridge’s Department of Theoretical and Applied Linguistics, and researchers from the Université de Montréal and New York ̽��ֱ��, Hill used the model as a way of bridging the gap between machines that understand the meanings of individual words and machines that can understand the meanings of phrases and sentences.

“Despite recent progress in AI, problems involving language understanding are particularly difficult, and our work suggests many possible applications of deep neural networks to language technology,” said Hill. “One of the biggest challenges in training computers to understand language is recreating the many rich and diverse information sources available to humans when they learn to speak and read.”

However, there is still a long way to go. For instance, when Hill’s system receives a query, the machine has no idea about the user’s intention or the wider context of why the question is being asked. Humans, on the other hand, can use their background knowledge and signals like body language to figure out the intent behind the query.

Hill describes recent progress in learning-based AI systems in terms of behaviourism and cognitivism: two movements in psychology that effect how one views learning and education. Behaviourism, as the name implies, looks at behaviour without looking at what the brain and neurons are doing, while cognitivism looks at the mental processes that underlie behaviour. Deep learning systems like the one built by Hill and his colleagues reflect a cognitivist approach, but for a system to have something approaching human intelligence, it would have to have a little of both.

“Our system can’t go too far beyond the dictionary data on which it was trained, but the ways in which it can are interesting, and make it a surprisingly robust question and answer system – and quite good at solving crossword puzzles,” said Hill. While it was not built with the purpose of solving crossword puzzles, the researchers found that it actually performed better than commercially-available products that are specifically engineered for the task.

Existing commercial crossword-answering applications function in a similar way to a Google search, with one system able to reference over 1100 dictionaries. While this approach has advantages if you want to look up a definition verbatim, it works less well when you input a question or query that the model has never seen in training. It also makes it incredibly ‘heavy’ in terms of the amount of memory it requires. “Traditional approaches are like lugging many heavy dictionaries around with you, whereas our neural system is incredibly light,” said Hill.

According to the researchers, the results show the effectiveness of definition-based training for developing models that understand phrases and sentences. They are currently looking at ways of enhancing their system, specifically by combining it with more behaviourist-style models of language learning and linguistic interaction.

Reference:
Hill, Felix et al. Learning to Understand Phrases by Embedding the Dictionary. Transactions of the Association for Computational Linguistics, [S.l.], v. 4, p. 17-30, feb. 2016. ISSN 2307-387X.

A web-based machine language system solves crossword puzzles far better than commercially-available products, and may help machines better understand language.

Despite recent progress in AI, problems involving language understanding are particularly difficult.

Felix Hill

Beth

Crossword

̽��ֱ��text in this work is licensed under a Creative Commons Attribution 4.0 International License. For image use please see separate credits above.

Yes

Licence type:

Attribution-Noncommerical