Rémi Monasson
CNRS & Ecole Normale Supérieure
Overview
Analytical Connectionism is a 2-week summer course on analytical tools, including methods from statistical physics and probability theory, for probing neural networks and higher-level cognition. The course brings together neuroscience, psychology and machine-learning communities, and introduces attendees to analytical methods for neural-network analysis and connectionist theories of higher-level cognition and psychology.
Connectionism, a key theoretical approach in psychology, uses neural-network models to simulate a wide range of phenomena, including perception, memory, decision-making, language, and cognitive control. However, most connectionist models remain, to a certain extent, black boxes, and we lack a mathematical understanding of their behaviors. Recent progress in theoretical neuroscience and machine learning has provided novel analytical tools that have advanced our mathematical understanding of deep neural networks, and have the potential to help make these “black boxes” more transparent.
During the School, teams of students work closely with faculty and postdoc mentors to develop research projects on topics related to analytical connectionism, presenting initial proposals during week one and interim results at the School’s conclusion. Participant projects from prior Schools have led to publications at NeurIPS.
Additionally, students are grouped based on their expertise and preferences and assigned to take notes for a specific lecturer. These notes are peer-reviewed and published in a special journal issue. Currently, lecture notes from the 2023 and 2024 editions of the school are being collected into a publication in the Proceedings of Machine Learning Research (PMLR). This initiative aims to make the content accessible to future participants and those who were unable to attend, while providing note-takers with the opportunity to contribute to a formal publication.
📚 This year's School will have a topical focus on "bias in learning."
What makes a learning system biased toward certain strategies? This year's School will examine case studies of deficits due to architectural or processing constraints alongside biases acquired as a consequence of statistical patterns in data.
In addition to this topical focus, attendees can expect to be introduced to tools and phenomena central to analytical connectionism (i.e., participants with no prior attendance of a School are encouraged to apply).
What makes a learning system biased toward certain strategies? This year's School will examine case studies of deficits due to architectural or processing constraints alongside biases acquired as a consequence of statistical patterns in data.
In addition to this topical focus, attendees can expect to be introduced to tools and phenomena central to analytical connectionism (i.e., participants with no prior attendance of a School are encouraged to apply).
This course will introduce:
- mathematical methods for neural-network analysis, providing a solid overview of the analytical tools available to understand neural-network models;
- key connectionist models with links to experimental observations, which provide targets for analytical results.
During the course, you will:
- attend lectures given by leading researchers on theoretical methods and applications, key connectionist models, and experimental observations;
- participate in tutorials, Q&A sessions, and panel discussions;
- present to and engage with lecturers, organizers, and other participants during a poster session;
- work in a group with other participants on a novel research project, mentored by the course organizers and lecturers.
Important dates
All dates are to be intended anywhere on earth time (AoE).
- Application deadline:
- April 18, 2025
- Outcome communicated:
- May 25, 2025
- Deadline to accept admission:
- June 15, 2025
Application details
Applications to participate in the 2025 School on Analytical Connectionism are now closed.
Target audience
This course is appropriate for graduate students, postdoctoral fellows and early-career faculty in a number of fields, including psychology, neuroscience, physics, computer science, and mathematics. Attendees are expected to have a strong background in one of these disciplines and to have made some effort to introduce themselves to a complementary discipline.
The course is limited to 40 attendees, who will be chosen to balance the representation of different fields. In circumstances where all other things are equal, priority will be given to applicants from underrepresented groups in STEM fields, using positive action under the UK Equality Act 2010 where appropriate.
Course fees
There is a course fee of 600 GBP. Attendees are expected to cover their own travel, accommodation and other subsistence expenses. Lunch and coffee/tea breaks will be provided on course days, and there will be one course dinner.
Cancellation policy:
- Within 7 days of registering/making the booking: 100% refund
- After 7 days of registering/making the booking and up to 15 days before the start of the school: 80% refund
- Less than 15 days before the start of the school: no refund
Financial assistance via travel grant may be available for successful applicants who find it difficult to take up a place for financial reasons. Applicants are asked to indicate in their application if they would like to be considered for financial aid. The amount of financial aid available will depend on the course funding from grants and sponsors.
Lecturers
Core lecturers
Topic lecturers
Course Content
This year, we are thrilled to have lecturers with expertise in the following research areas:
- analytical frameworks for understanding bias from statistical physics;
- perceptual and cognitive biases in human decision-making and belief formation;
- individual differences and deficits in learning, including atypical development and atypical cognitive processes;
- social learning and the emergence of group-level stereotypes;
- bias amplification in machine learning systems from theoretical and applied perspectives;
- clinical applications in computational psychiatry and mental health.
These components will be delivered as a set of core lectures in the first week, followed by a set of topic lectures in the second week.
Reading Material
Xuechunzi Bai
- Globally inaccurate stereotypes can result from locally adaptive exploration Bai, X., Fiske, S. T., & Griffiths, T. L. (2022). Psychological Science, 33(5), 671-684.
- Costly exploration produces stereotypes with dimensions of warmth and competence Bai, X., Griffiths, T. L., & Fiske, S. T. (2024). Journal of Experimental Psychology: General, 153(2), 347–357.
- Explicitly unbiased large language models still form biased associations Bai, X., Wang, A., Sucholutsky, I., & Griffiths, T. L. (2024). Proceedings of the National Academy of Sciences, 121(8), e2316228121.
- Aligned but Blind: Alignment Increases Implicit Bias by Reducing Awareness of Race Sun, L., Mao, C., Hofmann, V., & Bai, X. (2025). arXiv preprint arXiv:2506.00253.
Marco Baity-Jesi
- Deep Information Propagation Schoenholz, Samuel S., Justin Gilmer, Surya Ganguli, and Jascha Sohl-Dickstein. International Conference on Learning Representations (2017).
- On the Impact of the Activation Function on the Training Dynamics of Deep Neural Networks Soufiane Hayou, Arnaud Doucet, Judith Rousseau. Proceedings of the 36th International Conference on Machine Learning, PMLR 97 (2019): 2672-2680.
- A Theoretical Analysis of the Learning Dynamics under Class Imbalance Francazi, Emanuele, Marco Baity-Jesi, and Aurelien Lucchi. International Conference on Machine Learning. PMLR, 2023.
- Initial Guessing Bias: How Untrained Networks Favor Some Classes Francazi, Emanuele, Aurelien Lucchi, and Marco Baity-Jesi. arXiv preprint arXiv:2306.00809 (2023).
- Where You Place the Norm Matters: From Prejudiced to Neutral Initializations Emanuele Francazi, Francesco Pinto, Aurelien Lucchi, Marco Baity-Jesi. arXiv preprint arXiv:2505.11312 (2025).
- When the Left Foot Leads to the Right Path: Bridging Initial Prejudice and Trainability Alberto Bassi, Carlo Albert, Aurelien Lucchi, Marco Baity-Jesi, Emanuele Francazi. arXiv preprint arXiv:2505.12096 (2025).
Tiago Maia
- Reinforcement learning, conditioning, and the brain: Successes and challenges Maia, Tiago V. Cognitive, Affective, & Behavioral Neuroscience 9.4 (2009): 343-364.
- From reinforcement learning models to psychiatric and neurological disorders Maia, Tiago V., and Michael J. Frank. Nature neuroscience 14.2 (2011): 154-162.
- Changes in Corticostriatal Connectivity During Reinforcement Learning in Humans Horga, G., Maia, T.V., Marsh, R., Hao, X., Xu, D., Duan, Y., Tau, G.Z., Graniello, B., Wang, Z., Kangarlu, A. and Martinez, D., 2015. Human brain mapping, 36(2), pp.793-803.
- Chapter 10 of Computational psychiatry: A primer Vasco A. Conceição and Tiago V. Maia, edited by Seriès, Peggy, MIT Press, 2020.
Angela Radulescu
- Naturalistic reinforcement learning Wise, T., Emery, K., & Radulescu, A. (2024). Trends in Cognitive Sciences, 28(2), 144-158.
- Affect-congruent attention modulates generalized reward expectations Bennett, D., Radulescu, A., Zorowitz, S., Felso, V., & Niv, Y. (2023). PLoS computational biology, 19(12), e1011707.
- Modeling human eye movements during immersive visual search Radulescu, A., van Opheusden, B., Callaway, F., Griffiths, T. L., & Hillis, J. M. (2022). bioRxiv, 2022-12.
- Anxiety, avoidance, and sequential evaluation Zorowitz, S., Momennejad, I., & Daw, N. D. (2020). Computational Psychiatry (Cambridge, Mass.), 4, 10-1162.
- Dynamic interaction between reinforcement learning and attention in multidimensional environments Leong, Y. C., Radulescu, A., Daniel, R., DeWoskin, V., & Niv, Y. (2017). Neuron, 93(2), 451-463.
Tali Sharot
- Forming beliefs: Why valence matters. Sharot, T., & Garrett, N. (2016). Trends in cognitive sciences, 20(1), 25-33.
- Human reinforcement learning processes and biases: computational characterization and possible applications to behavioral public policy. Palminteri, S. (2025). Mind Soc (2025).
- Context-dependent outcome encoding in human reinforcement learning. Palminteri, S., & Lebreton, M. (2021). Current Opinion in Behavioral Sciences, 41, 144-151.
- Disinformation elicits learning biases. Vidal-Perez, J., Dolan, R. J., & Moran, R. (2025). eLife, 14.
- How people decide what they want to know. Sharot, T., & Sunstein, C. R. (2020). Nature Human Behaviour, 4(1), 14-19.
- How human–AI feedback loops alter human perceptual, emotional and social judgements. Glickman, M., & Sharot, T. (2025). Nature Human Behaviour, 9(2), 345-359.
- AI-induced hyper-learning in humans. Glickman, M., & Sharot, T. (2024). Current Opinion in Psychology, 60, 101900.
Fanny Yang
- Hidden yet quantifiable: A lower bound for confounding strength using randomized trials De Bartolomeis, P., Martinez, J.A., Donhauser, K. and Yang, F. In International Conference on Artificial Intelligence and Statistics 2024 (pp. 1045-1053). PMLR.
- Detecting critical treatment effect bias in small subgroups De Bartolomeis, P., Abad, J., Donhauser, K. and Yang, F. Conference on Uncertainty in Artificial Intelligence (UAI) 2024.
- Efficient Randomized Experiments Using Foundation Models De Bartolomeis, P., Abad, J., Wang, G., Donhauser, K., Duch, R. M., Yang, F., & Dahabreh, I. J. (2025). arXiv preprint arXiv:2502.04262.
Schedule
Participants
Contributed talks & posters
- Elaheh AKBARI-FATHKOUHI, “Distinct computational mechanisms underlying holistic processing of faces and line patterns”
- Federico BARRERA-LEMARCHAND, “Polarized crowds: The impact of linguistic complexity on group consensus”
- Anindita BASU, “Sticky thoughts: Effects of semantic topology and glassiness on the dynamics of memory transition”
- Sarah Kaarina CROCKFORD, “Using transformer-based word embedding similarity to model natural language usage”
- Ariane DELROCQ, “Towards a model of learning in the visual cortex”
- Yoav GER, “Learning dynamics of RNNs in closed-loop environments”
- Marcel GRAETZ, “Exploring parallel and sequential processing in RNNs with bottlenecks”
- Liz Aneth JARAMILLO HENAO, “Biologically plausible learning for spiking neural networks”
- Ishan KALBURGE, “Probabilistic representations fail to emerge in task-optimized networks”
- Alireza KARAMI, “Time‑resolved analysis of integer, fraction, and shape representations using MEG RSA”
- Clara KÜMPEL, “Heterosynaptic learning with structured Hessians”
- Amy X. LI, “Accounting for covert attention in value-guided choice”
- Matéo MAHAUT, “Differences of processing in vision models with universal representations”
- Siyuan MEI, “The asymmetric rotation response in ring attractors: How zebrafish compute head direction”
- Flavio NICOLETTI, “Statistical Mechanics of vector Hopfield networks near and above saturation”
- Monica PAOLETTI, “Integration of sensory evidence with rewards history in sequential decision making in humans and rats”
- Jacob PARKER, “Suboptimal human decision-making reflects an efficient information bottleneck on inference”
- Motahareh POURRAHIMI, “Emergent brain-like representations in a goal-directed neural network model of visual search”
- Yukun QU, “Development of the cognitive map predicts the emergence of human intelligence”
- Xiangjuan REN, “Effects of aging on memory and neural replay in humans”
- Joseph RUDOLER, “A framework for estimating the implicit bias of learning algorithms”
- Manoosh SAMIEI, “Multi-time scale reinforcement learning”
- Mandana SAMIEI, “The role of schemas in reinforcement learning: Insights and implications for generalization”
- Camilla SARRA, “Physics-inspired models for neural cell classification”
- Alice ZHANG, “From simple to complex: Shared learning dynamics in humans and neural networks”
- Stefania ZOI “Efficient Colour Coding in the Retina”
Participant list
- Elaheh AKBARI-FATHKOUHI
- Federico BARRERA-LEMARCHAND
- Anindita BASU
- Sarah Kaarina CROCKFORD
- Samuel DEBRAY
- Ariane DELROCQ
- Yoav GER
- Marcel GRAETZ
- Liz Aneth JARAMILLO HENAO
- Ishan KALBURGE
- Alireza KARAMI
- Clara KÜMPEL
- Adam LEE
- Amy X. LI
- Chenxiao MA
- Matéo MAHAUT
- Ingrid MARTIN
- Siyuan MEI
- Flavio NICOLETTI
- Monica PAOLETTI
- Jacob PARKER
- Motahareh POURRAHIMI
- Yukun QU
- Sepehr RAZAVI
- Xiangjuan REN
- Joseph RUDOLER
- Manoosh SAMIEI
- Mandana SAMIEI
- Camilla SARRA
- Janik SCHÜTTLER
- Alice ZHANG
- Stefania ZOI
Organizers
Course organizers
FAQ
- I need a visa to attend the programme. Will you provide the necessary documents for my visa application if I am accepted? If you require a UK Standard Visitor visa, we will provide an invitation letter to support your application once you confirm your acceptance into the programme. Please note that visa processing can take several weeks, so it’s necessary to start the application process as soon as you accept the offer.
- I am not based in London. Where can I stay during the programme? You may wish to look into UCL Summer Residences or University of London halls, both open to the general public during the student vacation period in summer. The YMCA Indian Student Hostel (if eligible) is within five minutes walking distance. London Metropolitan University also has a list of summer accommodation options on its website.
- Will I receive individual feedback if my application is not successful? Unfortunately, due to the large number of applicants, we are unable to offer individual feedback on unsuccessful applications.
- What are the requirements for poster presentations? Due to space constraints, posters must be in A1 portrait format (594mm x 841mm).
- What are the requirements for spotlight presentations? Presentations will be short, likely a maximum of 5 minutes. We will email final guidelines to all presenters closer to the programme start date.
- Other questions not covered in the FAQs? Contact us at admissions@gatsby.ucl.ac.uk.
Sponsors
This summer course is made possible by the generous support of the Gatsby Computational Neuroscience Unit (funded by the Gatsby Charitable Foundation), and the Sainsbury Wellcome Centre for Neural Circuits and Behaviour (funded by the Gatsby Charitable Foundation and the Wellcome Trust).
