
Thu 03 Jul 13:00: Changing Climate, Changing Corals: Predicting Long-Term Climatological Suitability for Tropical Reefs
Abstract
Stay tuned!
Bio
Orlando is a second-year PhD student with the AI4ER CDT . Supervised by Oscar Branson (Earth Sciences), he is interested in the opportunities and limitations for modelling marine ecosystems – particularly coral reefs – posed by the data available today.
- Speaker: Orlando Timmerman, University of Cambridge
- Thursday 03 July 2025, 13:00-14:00
- Venue: Room GS15 at the William Gates Building and on Zoom: https://cl-cam-ac-uk.zoom.us/j/4361570789?pwd=Nkl2T3ZLaTZwRm05bzRTOUUxY3Q4QT09&from=addon .
- Series: Energy and Environment Group, Department of CST; organiser: lyr24.
Thu 05 Jun 11:30: Unveiling complex transport processes in a large deep lake: From coastal upwelling to higher-mode internal waves
Abstract not available
- Speaker: Rafael Reiss (University of Cambridge)
- Thursday 05 June 2025, 11:30-12:30
- Venue: Open Plan Area, Institute for Energy and Environmental Flows, Madingley Rise CB3 0EZ.
- Series: Institute for Energy and Environmental Flows (IEEF); organiser: Catherine Pearson.
Fri 26 Sep 08:45: Title to be confirmed
Friday Morning Seminar Slot – CamVet Clinical Research Grants
- Speaker: Speaker to be confirmed
- Friday 26 September 2025, 08:45-10:00
- Venue: LT2.
- Series: Friday Morning Seminars, Dept of Veterinary Medicine; organiser: Fiona Roby.
Tue 03 Jun 16:00: Computational wireless sensing for health Zoom: https://cam-ac-uk.zoom.us/j/82442600092?pwd=DfDLE12p2Kmnh532rSMVB8mOn7uLDa.1
Abstract: Mobile computing technologies have advanced substantially over the last two decades. Today, the smart devices enabled by economies of scale, incorporate high-quality wireless sensors such as acoustic and RF sensors and the trend shows an increase in both the quantity and quality of these sensors. These sensors can be leveraged to enable a contactless passive monitoring of physiological signals of subjects and early diagnoses of various health conditions. In this talk, I will present a privacy aware wireless sensing technology to enable equitable, passive contactless monitoring of breathing and heart rate signals to detect opioid overdose in a timely manner.
Bio: Rajalakshmi Nandakumar is an Assistant Professor at the Jacobs Technion-Cornell Institute at Cornell Tech and in the Information Science department at Cornell University. She received her Ph.D. from University of Washington in Computer Science and Engineering in 2019. Her research focuses on developing wireless sensing technologies that enable novel applications in various domains including mobile health, user interfaces and IoT networks. She developed the first contactless smartphone-based sleep apnea diagnosis system that was licensed by ResMed Inc. and now used by millions of users for sleep staging. She was recognized with the UW Medicine Judy Su Clinical Research award, Paul Baran Young Scholar award by the Marconi Society and also named as the rising star in EECS by MIT .
Zoom: https://cam-ac-uk.zoom.us/j/82442600092?pwd=DfDLE12p2Kmnh532rSMVB8mOn7uLDa.1
- Speaker: Rajalakshmi Nandakumar, Cornell University
- Tuesday 03 June 2025, 16:00-17:00
- Venue: Online.
- Series: Mobile and Wearable Health Seminar Series; organiser: Cecilia Mascolo.
Fri 06 Jun 14:00: Complexity of sampling truncated log-concave measures, and the role of stochastic localization
Motivated by computational challenges in Bayesian models with indicator variables, such as probit/tobit regression, we study the computational complexity of drawing samples from a truncated log-concave measure. We discuss two problems. In the first part, using stochastic localization as a way to reduce the sampling problem to truncated Gaussians, we analyze the hit-and-run algorithm for sampling uniformly from an isotropic convex body in n dimensions and establish $n2$ mixing time. In the second part, building on interior point methods, we analyze the mixing time of regularized Dikin walks for sampling log-concave measures truncated on a polytope. For a logconcave and log-smooth distribution with condition number $\kappa$, truncated on a polytope in $Rn$ defined with $m$ linear constraints, we prove that the soft-threshold Dikin walk mixes in $O((m+\kappa)n)$ iterations from a warm initialization. It improves upon prior work which required the polytope to be bounded and involved a bound dependent on the radius of the bounded region. Here, stochastic localization allows us to extend the analysis to weakly log-concave measures. https://arxiv.org/abs/2212.00297 https://arxiv.org/abs/2412.11303
- Speaker: Yuansi Chen (ETH Zurich)
- Friday 06 June 2025, 14:00-15:00
- Venue: MR12, Centre for Mathematical Sciences.
- Series: Statistics; organiser: Qingyuan Zhao.
Wed 11 Jun 16:00: Covert cnidarians: cryptic lives of the endoparasitic Myxozoa Host: Juliana Naldoni
Myxozoans are a diverse clade of endoparasites with complex life cycles and are the causative agents of some devastating fish diseases. Their phylogenetic placement was long obscure due to extreme morphological simplification and rapid evolution, but they are now established as a radiation of endoparasitic cnidarians that exploit freshwater, marine and terrestrial hosts. I will review diversity, lifestyles, and morphological simplification that characterise these generally unfamiliar animals and then present insights on how myxozoans exploit their invertebrate hosts and disperse to colonise new freshwater environments. By so revealing the cryptic lives of myxozoans we can appreciate how particular cnidarian traits may have facilitated and promoted this remarkable endoparasitic radiation.
Host: Juliana Naldoni
- Speaker: Prof Beth Okamura, Natural History Museum
- Wednesday 11 June 2025, 16:00-17:00
- Venue: Lecture Theater, Department of Pathology, Tennis Court Road.
- Series: Parasitology Seminars; organiser: Ross Waller.
Thu 05 Jun 13:00: Towards Improved Crop Type Classification: a Compact Representation Approach for Smallholder Agriculture (TESSERA application)
Abstract
Satellite-based monitoring of smallholder agriculture is an important tool for food security but existing approaches are neither accessible nor effective for small plot field systems. To address these issues, crop type classification using representations generated by a global foundation model, TESSERA , is compared to best classification approaches in the literature. We present a novel approach to smallholder plots and compare representation based methods to raw data based methods for crop type classification in challenging environments. We find that our representation based approach offers a triple win: 1) consistent and statistically significant performance improvement over current methods, 2) greater simplicity due to the elimination of cloud masking and feature engineering, and 3) the reduction of computational cost. Our representation based approach achieves significantly higher F1 scores in the classification of 7 crop types for small fields in Austria for 5 classes (over 10% improvement in one case) and comparable F1 scores for two classes, and the best representation-based methods use 5% and 8% of compute compared to the best raw data method. These results indicate that representations are an effective approach for crop type classification tasks for small field systems.
Bio
Madeline Lisaius received BS and MS degrees in Earth Systems with a focus on environmental spatial statistics and remote sensing from Stanford University, Stanford, California, USA as well as MRes degree in Environmental Data Science from the University of Cambridge, Cambridge, UK. She is working towards the PhD in the Department of Computer Science and Technology at the University of Cambridge. She is focused on topics of food security and environmental justice, remote sensing, and machine learning.
- Speaker: Madeline Lisaius, University of Cambridge
- Thursday 05 June 2025, 13:00-14:00
- Venue: Room GS15 at the William Gates Building and on Zoom: https://cl-cam-ac-uk.zoom.us/j/4361570789?pwd=Nkl2T3ZLaTZwRm05bzRTOUUxY3Q4QT09&from=addon .
- Series: Energy and Environment Group, Department of CST; organiser: lyr24.
Fri 06 Jun 16:00: Numerical simulations of multiphase flows with various complexities
Multiphase flows are of central importance to a wide range of industrial applications and environmental settings. Examples of these include mixing in stirred vessels and static mixers, flows in micro-channels and microfluidics devices, falling films for CO2 capture, and aerosol formation via bubble bursting through interfaces in the oceans. Some of these flows feature the presence of surface-active agents (surfactants), present either by design or as contaminants. Furthermore, multiphase flows are often punctuated by topological transitions related to the coalescence of dispersed drops or bubbles, and the breakup of threads or ligaments. Here, we provide a few examples of interest to the JFM community but focus on drop impact on hydrophobic substrates in the presence of surfactants above the critical micelle concentration. Our model accounts for the spatio-temporal evolution of the surfactants along the interface and within the bulk; the bulk and interfacial species are fully-coupled via sorptive fluxes. Micellar formation and breakup are also accounted for, and the surfactant dynamics are coupled to the flow through the dependence of the surface tension on the local interfacial surfactant concentration. Our numerical procedure is based on the use of a hybrid interface-tracking/level-set approach. The results of our parametric study help identify the various physical mechanisms underlying the observed flow phenomena.
- Speaker: Prof Omar Matar, Imperial College London
- Friday 06 June 2025, 16:00-17:00
- Venue: https://cassyni.com/s/fmws.
- Series: Fluid Mechanics (DAMTP); organiser: Professor Grae Worster.
Tue 03 Jun 11:00: Discovering reward-guided learning strategies from large-scale datasets
Understanding the neural mechanisms of reward-guided learning is a long-standing goal of computational neuroscience. Recent methodological innovations enable us to collect ever larger neural and behavioral datasets. This presents opportunities to achieve greater understanding of learning in the brain at scale, as well as methodological challenges. In the first part of the talk, I will discuss our recent insights into the mechanisms by which zebra finch songbirds learn to sing. Dopamine has been long thought to guide reward-based trial-and-error learning by encoding reward prediction errors. However, it is unknown whether the learning of natural behaviours, such as developmental vocal learning, occurs through dopamine-based reinforcement. Longitudinal recordings of dopamine and bird songs reveal that dopamine activity is indeed consistent with encoding a reward prediction error during naturalistic learning.
In the second part of the talk, I will talk about recent work we are doing at DeepMind to develop tools for automatically discovering interpretable models of behavior directly from animal choice data. Our method, dubbed CogFunSearch, uses LLMs within an evolutionary search process in order to “discover” novel models in the form of Python programs that excel at accurately predicting animal behavior during reward-guided learning. The discovered programs reveal novel patterns of learning and choice behavior that update our understanding of how the brain solves reinforcement learning problems.
- Speaker: Kimberly Stachenfeld (DeepMind, Columbia)
- Tuesday 03 June 2025, 11:00-12:30
- Venue: CBL Seminar Room, Engineering Department, 4th floor Baker building.
- Series: Computational Neuroscience; organiser: Daniel Kornai.
Tue 03 Jun 13:30: When is a state a belief state?
When making predictions under uncertainty, neural networks often exhibit behaviors reminiscent of probabilistic inference—even without explicit inductive biases. In reinforcement learning (RL), uncertainty about latent variables is naturally captured by Partially Observable Markov Decision Processes (POMDPs). These models apply broadly, from games like poker to domains such as healthcare and autonomous driving.
In POMD Ps, the optimal agent performs sequential Bayesian filtering under a generative model to infer the latent state of the environment—maintaining a “belief” over that state. Surprisingly, naively trained recurrent neural networks (RNNs) can outperform dedicated probabilistic inference methods tailored to POMD Ps. Recent work has shown that RNNs trained to learn value functions can develop belief-like representations without access to the generative model—though lacking theoretical justification for this phenomenon.
In this talk, I will connect these observations by presenting theoretical arguments for when and why belief-like representations emerge in deep RL agents. I will also demonstrate how such theoretical insights can inform and justify auxiliary loss objectives used in state-of-the-art architectures, such as DreamerV3.
- Speaker: Daniel Kornai (CBL)
- Tuesday 03 June 2025, 13:30-15:00
- Venue: CBL Seminar Room, Engineering Department, 4th floor Baker building.
- Series: Computational Neuroscience; organiser: Daniel Kornai.
Tue 03 Jun 13:30: When is a state a belief state?
When making predictions under uncertainty, neural networks often exhibit behaviors reminiscent of probabilistic inference—even without explicit inductive biases. In reinforcement learning (RL), uncertainty about latent variables is naturally captured by Partially Observable Markov Decision Processes (POMDPs). These models apply broadly, from games like poker to domains such as healthcare and autonomous driving.
In POMD Ps, the optimal agent performs sequential Bayesian filtering under a generative model to infer the latent state of the environment—maintaining a “belief” over that state. Surprisingly, naively trained recurrent neural networks (RNNs) can outperform dedicated probabilistic inference methods tailored to POMD Ps. Recent work has shown that RNNs trained to learn value functions can develop belief-like representations without access to the generative model—though lacking theoretical justification for this phenomenon.
In this talk, I will connect these observations by presenting theoretical arguments for when and why belief-like representations emerge in deep RL agents. I will also demonstrate how such theoretical insights can inform and justify auxiliary loss objectives used in state-of-the-art architectures, such as DreamerV3.
- Speaker: Daniel Kornai (CBL)
- Tuesday 03 June 2025, 13:30-15:00
- Venue: CBL Seminar Room, Engineering Department, 4th floor Baker building.
- Series: Computational Neuroscience; organiser: Daniel Kornai.
Tue 02 Jun 13:30: When is a state a belief state?
When making predictions under uncertainty, neural networks often exhibit behaviors reminiscent of probabilistic inference—even without explicit inductive biases. In reinforcement learning (RL), uncertainty about latent variables is naturally captured by Partially Observable Markov Decision Processes (POMDPs). These models apply broadly, from games like poker to domains such as healthcare and autonomous driving.
In POMD Ps, the optimal agent performs sequential Bayesian filtering under a generative model to infer the latent state of the environment—maintaining a “belief” over that state. Surprisingly, naively trained recurrent neural networks (RNNs) can outperform dedicated probabilistic inference methods tailored to POMD Ps. Recent work has shown that RNNs trained to learn value functions can develop belief-like representations without access to the generative model—though lacking theoretical justification for this phenomenon.
In this talk, I will connect these observations by presenting theoretical arguments for when and why belief-like representations emerge in deep RL agents. I will also demonstrate how such theoretical insights can inform and justify auxiliary loss objectives used in state-of-the-art architectures, such as DreamerV3.
- Speaker: Daniel Kornai (CBL)
- Tuesday 02 June 2026, 13:30-15:00
- Venue: CBL Seminar Room, Engineering Department, 4th floor Baker building.
- Series: Computational Neuroscience; organiser: Daniel Kornai.
Wed 04 Jun 13:00: The Australian Antarctic Program Partnership (AAPP) Biogeochemistry Project: Understanding the changing Southern Ocean carbon cycle
The Australian Antarctic Program Partnership (AAPP) is focused on understanding the nature and impacts of Southern Ocean Change. The Biogeochemistry Project, one of the seven complementary initiatives within the AAPP , combines observations, models and data syntheses to understand changes in the Southern Ocean carbon cycle. This work is undertaken in collaboration with other government agencies, national infrastructure programs, and academic institutions, and highlights the use of essential ocean observations and models to improve understanding and deliver impact. An overview of recent field programs will be presented, along with new work to quantify the uptake and storage of anthropogenic CO2 in the ocean, to validate estimates of ocean carbon export from autonomous platforms, and to improve model representation of air-sea CO2 exchange.
- Speaker: Elizabeth H. Shadwick, CSIRO Environment
- Wednesday 04 June 2025, 13:00-14:00
- Venue: BAS Seminar Room 1.
- Series: British Antarctic Survey - Polar Oceans seminar series; organiser: Dr Birgit Rogalla.
Wed 04 Jun 11:00: Exploring charge density waves in twisted bilayer NbSe2 with machine learning
This is a joint seminar with the LJC .
Niobium diselenide has garnered significant attention over the past few decades because of the coexis tence of superconductivity and charge density waves (CDWs), observable down to the monolayer limit. Introducing relative twist angles between monolayers, in the field of twistronics, offers a new variable to tune these systems, yet a fundamental question remains: do CDWs persist in moiré structures, and how are they altered compared to the pristine monolayer/bilayer? Traditional first-principles methods face limitations due to the computational resources required for long-wavelength moiré patterns; for instance, a 1-degree twist angle necessitates modeling over 10,000 atoms, making simulations impractical. This study employs first-principles data to develop machine learning interatomic potentials with the Allegro architecture, enabling scalable and accurate simulations. We investigate the formation and evolution of CDW order in monolayers and twisted bilayers, validating our results against density functional theory calculations with minimal errors in energy and forces. Beyond niobium diselenide, our goal is to establish a protocol for studying CDWs in two-dimensional systems. We outline strategies for producing training data and perform a detailed hyperparameter scan to identify key aspects for studying these systems [1].
- Norma Rivano et al. arXiv.2504.13675 2025
- Speaker: Dr Zac Goodwin (Oxford)
- Wednesday 04 June 2025, 11:00-12:00
- Venue: Seminar Room 3, RDC.
- Series: Theory of Condensed Matter; organiser: Bo Peng.
Thu 05 Jun 14:00: “Old and new physics in the J1-J2 Heisenberg spin chain”
In this talk, I will briefly describe two of our recent pieces of work on the physics of the J1-J2 Heisenberg spin chain, a one-dimensional array of quantum spins coupled by first- and second-neighbour exchange interactions. The first piece of work [1] concerns the case where J1 and J2 are both antiferromagnetic: in this case there is a phase transition from a Luttinger liquid to a valence bond solid as J2/J1 is increased, and we provide a novel direct method to derive the field theory that describes the critical point between these two phases. The second piece [2] concerns the case where J1 is antiferromagnetic but J2 is strongly ferromagnetic: counter-intuitively, there is a transition in this case as well, but this time of a ‘liquid-to-liquid’ type. We present a field-theory description of it, and an analogue system of three coupled chains that helps to illustrate the physics.
[1] F. Azad, A. J. McRoberts, CAH , and A. G. Green, “Generalized Haldane map from the matrix product state path integral to the critical theory of the J1-J2 chain,” Phys. Rev. Research 7, L012037 (2025).
[2] A. J. McRoberts, CAH , and A. G. Green, “Transition between critical antiferromagnetic phases in the J1-J2 spin chain,” arXiv:2411.08095v2 (2025).
- Speaker: Prof. Chris Hooley (Coventry)
- Thursday 05 June 2025, 14:00-15:30
- Venue: Seminar Room 3, RDC.
- Series: Theory of Condensed Matter; organiser: Bo Peng.
Tue 21 Oct 11:15: Title TBC
Abstract TBC
- Speaker: Dr. Weiyang Wang (University of Chinese Academy of Sciences)
- Tuesday 21 October 2025, 11:15-12:00
- Venue: TBC.
- Series: Hills Coffee Talks; organiser: Charles Walker.
Tue 01 Jul 11:15: Title TBC
Abstract TBC
- Speaker: Prof. Howard Reader
- Tuesday 01 July 2025, 11:15-12:00
- Venue: Coffee area, Battcock Centre.
- Series: Hills Coffee Talks; organiser: Charles Walker.
Wed 11 Jun 16:00: TBA Host: Juliana Naldoni
Abstract not available
Host: Juliana Naldoni
- Speaker: Prof Beth Okamura, Natural History Museum
- Wednesday 11 June 2025, 16:00-17:00
- Venue: Lecture Theater, Department of Pathology, Tennis Court Road.
- Series: Parasitology Seminars; organiser: jn472.
Wed 11 Jun 13:30: Title tbc
Abstract tbc
- Speaker: Davi de Castro Silva (University of Cambridge)
- Wednesday 11 June 2025, 13:30-15:00
- Venue: MR4, CMS.
- Series: Discrete Analysis Seminar; organiser: Julia Wolf.
Tue 03 Jun 11:15: Prebiotic Chemistry, Exoplanets and Stellar Flaring
Nitroprusside is an important prebiotic molecule, thought to contribute to reaction pathways that lead to the production of amino acid chains (Mariani et al. [2018]). Nitroprusside can be made from Ferrocyanide photochemically. It has been found that the timescales for this reaction on Early Earth would have been between an order of days to months , making this route of abiotic production very useful in further prebiotic reaction networks and an important factor to consider when discussing the viability of life to evolve on a planet (Rimmer et al. [2021]). Here we investigate this reaction with a focus on constant and time varied radiation, meaning experimental runs involving the sample being subjected to a constant flux of UV light and runs with UV flux changing over time. FlareLab makes use of a broad band UV-Vis Laser Driven Light Source (LDLS), to experimentally simulate stellar irradiation and stellar flaring activity. The reasoning behind investigating flares is based on recent findings that have shown that M-dwarves are prone to flaring (G¨unther et al. [2020]). Flaring for M-dwarves is also shown to be the best way to get enough UV to an exoplanet’s surface for good yield of photochemical products (Ranjan et al. [2017]). With M-dwarves seen as the best stars to look at to detect small rocky planets, it is important to consider how flaring could effect the production of Nitroprusside and if there’s a discrepancy between assuming a constant irradiation of the surface or taking into account flaring.
We show that FlareLab can be used as a means of detecting the production of Nitroprusside in-situ during the irradiation period. We also compare the constant flux and variable flux regimes, and discuss the implications of these findings.
- Speaker: Lukas Rossmanith
- Tuesday 03 June 2025, 11:15-12:00
- Venue: Martin Ryle Seminar Room, Kavli Institute.
- Series: Hills Coffee Talks; organiser: David Buscher.