A core goal of the lab is to make machines learn more like human infants, and here is a summary of some of our recent work in this area.
Developmental psychologists have shown that infants are learning many things in their first year. However, linguistic understanding is primitive until the end of the year, and so their learning must be "self-supervised", in that they can learn without being explicitly taught. At present, machines are mostly taught using hand-curated datasets, which are painstakingly labelled by humans. Self-supervised learning algorithms can potentially reduce the dependence on these datasets, and so are of great interest to the machine learning community.
In an arXiv preprint, Lorijn Zaadnoordijk from the lab, and our collaborator Tarek Besold have reviewed the developmental psychology literature to identify potential "next big thing(s)" for this area of machine learning.
Humans have a deep understanding of the world. When we recognise an object, we know what other things it is similar to and we can classify it as part of some superordinate category. This type of knowledge is called semantic knowledge. Cliona O'Doherty has been testing the idea that by observing the co-occurrences of objects in the world, infants could not just learn how to recognise things, but also learn about semantics. She has done this by setting up a computational model using a deep neural network.
Cliona O'Doherty will present SemanticCMC - improved semantic self-supervised learning with naturalistic temporal co-occurrences at the workshop Self-supervised learning: theory and practice at Neural Information Processing Systems (NeurIPS) 2020.
How Can Random Networks Explain the Brain So Well?
A part of the brain called the inferotemporal (IT) cortex is critical for humans and other monkeys to visually recognise objects. Currently, deep neural networks are the best models of brain responses in the IT cortex of adults. It has been argued that this is because the visual features that deep neural networks learn for object recognition are the same as those IT uses. But, Anna Truzzi has been investigating a conundrum, which is that actually untrained (or random) deep neural networks also do a surprisingly good job of modelling IT activity.
Anna presented the paper "Convolutional Neural Networks as a Model of Visual Activity in The Brain: Greater Contribution of Architecture Than Learned Weights" at the workshop Bridging AI and Cognitive Science at the International Conference on Learning Representations (ICLR) 2020. She will also be presenting at the NeurIPS2020 workshop Shared Visual Representations in Humans and Machine Intelligence, with the title "Understanding CNNs as a model of the inferior temporal cortex: using mediation analysis to unpack the contribution of perceptual and semantic features in random and trained networks". This work is also directly relevant to neuroscientists, and was presented at the neuromatch 1.0 conference with the title "Are deep neural networks effective models of visual activity in the brain because of their architecture or training?".
This was a fun project! We tried using an off-the-shelf face recognition tool, Amazon Rekognition, to score infant video data collected online.
Chouinard, B., Scott, K., Cusack, R. (in press). Using Automatic Face Analysis to Score Infant Behaviour from Video Collected Online. Infant Behavior and Development. preprint
The lab's penultimate student from Western University, Patrick Gatutsi, receiving his MSc in Neuroscience.
Congratulations to Patrick Gatutsi, who passed is Master's defense with minor changes. He worked to develop a wearable device to measure fetal movements in response to sound, and conducted extensive evaluation in Rwanda.
It was great to see past and present PhD students from the lab in three different countries in Singapore for the Human Brain Mapping conference, along with a potential member of the next generation of neuroimagers. Thank you to everyone that came to Laura Cabral's and Chiara Caldinelli's posters. And, to Chiara for her work on the committee of the Student & postdoc special interest group.
Thanks to the organisers of the International Congress in Infant Studies in Philadelphia. Laura Cabral and I had a great time at the meeting.
Many members of the lab have published work that I'm very excited about this year. Some of the publications in 2018 include:
We are excited to announce the ERC "Foundations of Cognition" project, which will begin in October 2018. Read the launch summary.
I was honoured to be featured in the Provost's Annual Review of 2016-2017 along with the other professors appointed at Trinity this year. The review also gives an overview of Trinity in 2017, including the entrance to the League of European Research Universities, about student experience and public engagement.
It has been a busy summer for the lab. In addition to getting set up at Trinity College Dublin, we've been to some fantastic conferences and met some fascinating people.
We're excited to have been selected as organizers of a symposium at the Human Brain Mapping meeting this June. "Collect Your Thoughts: Individual Differences in the Networks Underlying Intelligence," examines how information is brought together to allow the complex, flexible cognition that is fundamental to human intelligence.
I'm excited to be working with a fantastic group of presenters representing four countries and fittingly for International Women's Day, equally representing the genders:
Rhodri Cusack (Trinity College Dublin, Ireland) "The Fronto-Parietal Network is Maturely Connected and Influences Developing Behaviour from the First Months."
Rogier Kievit (University of Cambridge, UK) "Watershed Models of Intelligence Through the Lifespan"
Danielle Bassett (University of Pennsylvania, USA) “Charting dynamic interactions between large-scale brain networks in health and disease”
Lorina Naci (Western University, Canada) “The Neural Machinery of Conscious Cognition: Converging Evidence from Anesthesia-Induced Unconsciousness, Severe Brain Injury and Intellectual Prowess.”
Applications are invited for a PhD scholarship in the rapidly expanding field of infant neuroimaging. The project’s goals are to understand how cognitive functions emerge, and to address the pressing clinical problem of detecting which infants with perinatal brain injury will develop atypical cognitive function to facilitate focused and timely intervention.
The project will characterize the development of brain systems in the first year after birth using magnetic resonance imaging (MRI). It will require an interdisciplinary synthesis across fields, including models of brain function from cognitive neuroscience, MRI acquisition, and data analysis with machine learning. Experience in each of these areas, or a willingness and aptitude to learn them, is essential. You will hold a first or upper second class honours degree (or equivalent) in psychology, computer science, neuroscience, physics, or a related field.
You will be based at the prestigious Trinity College Institute of Neuroscience, on Trinity College’s campus in the heart of Dublin, which houses 3T and 7T MRI scanners and has strong clinical collaborations. The successful candidate will join an exciting and dynamic research team and will be encouraged to develop their knowledge, technical skills and transferable skills.
The PhD will be directly supervised by Rhodri Cusack, the incoming Thomas Mitchell Professor of Cognitive Neuroscience.
To apply, please send before April 1, 2017, the following items by email to rhodri [at] cusacklab.org with the subject line PHD17:
* a cover letter explaining why this project interest you and what you will bring to it
* your curriculum vitae
* your transcript or grades (unofficial is fine at this stage)
* the names of three referees with email addresses and phone numbers
For further details:
* Cusack laboratory http://www.cusacklab.org
* Trinity College Institute of Neuroscience http://www.tcd.ie/Neuroscience
* Trinity College http://www.tcd.ie (video: http://www.youtube.com/watch?v=J8evbCLVepg)