Thu 21 Aug 12:45: GNN-ACLP: Graph Neural Networks Based Analog Circuit Link Prediction
Circuit link prediction identifying missing component connections from incomplete netlists is crucial in analog circuit design automation. However, existing methods face three main challenges: 1) Insufficient use of topological patterns in circuit graphs reduces prediction accuracy; 2) Data scarcity due to the complexity of annotations hinders model generalization; 3) Limited adaptability to various netlist formats. We propose GNN -ACLP, a graph neural networks (GNNs) based method featuring three innovations to tackle these challenges. First, we introduce the SEAL (learning from Subgraphs, Embeddings, and Attributes for Link prediction) framework and achieve port-level accuracy in circuit link prediction. Second, we propose Netlist Babel Fish, a netlist format conversion tool leveraging retrieval-augmented generation (RAG) with a large language model (LLM) to improve the compatibility of netlist formats. Finally, we construct SpiceNetlist, a comprehensive dataset that contains 775 annotated circuits across 10 different component classes. Experiments demonstrate accuracy improvements of 16.08% on SpiceNetlist, 11.38% on Image2Net, and 16.01% on Masala-CHAI compared to the baseline in intra-dataset evaluation, while maintaining accuracy from 92.05% to 99.07% in cross-dataset evaluation, exhibiting robust feature transfer capabilities.
- Speaker: Guanyuan Peter Pan
- Thursday 21 August 2025, 12:45-13:45
- Venue: Zoom (Meeting ID: 873 5011 9733, Passcode: 732177).
- Series: Foundation AI; organiser: Pietro Lio.
Thu 21 Aug 12:30: GNN-ACLP: Graph Neural Networks Based Analog Circuit Link Prediction
Circuit link prediction identifying missing component connections from incomplete netlists is crucial in analog circuit design automation. However, existing methods face three main challenges: 1) Insufficient use of topological patterns in circuit graphs reduces prediction accuracy; 2) Data scarcity due to the complexity of annotations hinders model generalization; 3) Limited adaptability to various netlist formats. We propose GNN -ACLP, a graph neural networks (GNNs) based method featuring three innovations to tackle these challenges. First, we introduce the SEAL (learning from Subgraphs, Embeddings, and Attributes for Link prediction) framework and achieve port-level accuracy in circuit link prediction. Second, we propose Netlist Babel Fish, a netlist format conversion tool leveraging retrieval-augmented generation (RAG) with a large language model (LLM) to improve the compatibility of netlist formats. Finally, we construct SpiceNetlist, a comprehensive dataset that contains 775 annotated circuits across 10 different component classes. Experiments demonstrate accuracy improvements of 16.08% on SpiceNetlist, 11.38% on Image2Net, and 16.01% on Masala-CHAI compared to the baseline in intra-dataset evaluation, while maintaining accuracy from 92.05% to 99.07% in cross-dataset evaluation, exhibiting robust feature transfer capabilities.
- Speaker: Guanyuan Peter Pan
- Thursday 21 August 2025, 12:30-13:30
- Venue: Zoom (Meeting ID: 873 5011 9733, Passcode: 732177).
- Series: Foundation AI; organiser: Pietro Lio.
Fri 14 Nov 12:00: Title to be confirmed
Abstract not available
- Speaker: Kris Cao (Cohere)
- Friday 14 November 2025, 12:00-13:00
- Venue: Hybrid (In-Person + Online). Here is the Zoom link: https://cam-ac-uk.zoom.us/j/4751389294?pwd=Z2ZOSDk0eG1wZldVWG1GVVhrTzFIZz09.
- Series: NLIP Seminar Series; organiser: Suchir Salhan.
Fri 14 Nov 12:00: Title to be confirmed
Abstract not available
- Speaker: Kris Cao (Cohere)
- Friday 14 November 2025, 12:00-13:00
- Venue: Hybrid (In-Person + Online). Here is the Zoom link: https://cam-ac-uk.zoom.us/j/4751389294?pwd=Z2ZOSDk0eG1wZldVWG1GVVhrTzFIZz09.
- Series: NLIP Seminar Series; organiser: Suchir Salhan.
Mon 15 Sep 13:00: Multi-modal modeling in precision medicine: from data imputation to synthetic data
Missing data presents a persistent challenge in biomedical research. Data imputation techniques have evolved from single-modality approaches to multi-modal approaches, which show great promise for imputing one modality based on the availability of another. Recent advancements in large, pre-trained artificial intelligence (AI) models, known as foundation models, offer even more powerful solutions for data imputation. We introduce the concept of cross-modal data modeling, a methodology harnessing foundation models to impute missing data and also generate realistic synthetic samples. Multi-modal modeling empowers researchers to model complex interactions among diverse biomedical data types, including omics and imaging. This approach can illuminate how one modality influences another, facilitating in-silico exploration of disease mechanisms without the need for extensive and costly real-world data collection. We highlight ongoing efforts in multi-modal modeling in spatial omics, digital pathology and radiology, and anticipate its substantial contributions to understanding disease biology and enhancing healthcare practices.
- Speaker: Dr. Olivier Gevaert - Associate Professor, Depts of Medicine & Biomedical Data Science, Stanford University
- Monday 15 September 2025, 13:00-14:00
- Venue: CRUK CI Lecture Theatre.
- Series: Seminars on Quantitative Biology @ CRUK Cambridge Institute ; organiser: Simona Valeviciute.
Fri 10 Oct 08:45: ALI Culture of Nasal Epithelial Cells for Investigating Inflammatory Responses
Abstract not available
- Speaker: Julieth Murillo Silva, Department of Veterinary Medicine
- Friday 10 October 2025, 08:45-10:00
- Venue: LT2.
- Series: Friday Morning Seminars, Dept of Veterinary Medicine; organiser: Fiona Roby.
Thu 21 Aug 12:30: GNN-ACLP: Graph Neural Networks Based Analog Circuit Link Prediction
Circuit link prediction identifying missing component connections from incomplete netlists is crucial in analog circuit design automation. However, existing methods face three main challenges: 1) Insufficient use of topological patterns in circuit graphs reduces prediction accuracy; 2) Data scarcity due to the complexity of annotations hinders model generalization; 3) Limited adaptability to various netlist formats. We propose GNN -ACLP, a graph neural networks (GNNs) based method featuring three innovations to tackle these challenges. First, we introduce the SEAL (learning from Subgraphs, Embeddings, and Attributes for Link prediction) framework and achieve port-level accuracy in circuit link prediction. Second, we propose Netlist Babel Fish, a netlist format conversion tool leveraging retrieval-augmented generation (RAG) with a large language model (LLM) to improve the compatibility of netlist formats. Finally, we construct SpiceNetlist, a comprehensive dataset that contains 775 annotated circuits across 10 different component classes. Experiments demonstrate accuracy improvements of 16.08% on SpiceNetlist, 11.38% on Image2Net, and 16.01% on Masala-CHAI compared to the baseline in intra-dataset evaluation, while maintaining accuracy from 92.05% to 99.07% in cross-dataset evaluation, exhibiting robust feature transfer capabilities.
- Speaker: Guanyuan Peter Pan
- Thursday 21 August 2025, 12:30-13:30
- Venue: Zoom (details TBA).
- Series: Foundation AI; organiser: Pietro Lio.
Thu 16 Oct 14:00: Title to be confirmed
Abstract not available
- Speaker: Arijeet Pal, UCL
- Thursday 16 October 2025, 14:00-15:15
- Venue: Seminar Room 3, RDC.
- Series: Theory of Condensed Matter; organiser: Gaurav.
Mon 02 Feb 18:00: Emergent laws in structural vibration with application to the design of engineering systems
One of the many outstanding achievements of G I Taylor was the discovery of relatively simple statistical laws that apply to highly complex turbulent flows. The emergence of simple laws from complexity is well known in other branches of physics, for example the emergence of the laws of heat conduction from molecular dynamics. Complexity can also arise at large scales, and the structural vibration of an aircraft or a car can be a surprisingly difficult phenomenon to analyse, partly because millions of degrees of freedom may be involved, and partly because the vibration can be extremely sensitive to small changes or imperfections in the system. In this talk it is shown that the prediction of vibration levels can be much simplified by the derivation and exploitation of emergent laws, analogous to some extent to the heat conduction equations, but with an added statistical aspect, as in turbulent flow. The emergent laws are discussed and their application to the design of aerospace, marine, and automotive structures is described. As an aside it will be shown that the same emergent theory can be applied to a range of problems involving electromagnetic fields.
- Speaker: Emeritus Professor of Mechanical Engineering, Department of Engineering, University of Cambridge
- Monday 02 February 2026, 18:00-19:00
- Venue: Bristol-Myers Squibb Lecture Theatre, Department of Chemistry.
- Series: Cambridge Philosophical Society; organiser: Beverley Larner.
Mon 02 Mar 18:00: What is Digital Identity all about? - Professor Jon Crowcroft
We have many forms of identity, whether socially constructed (kinship, personas, relationships), or issued via organisations (employers, banks, clubs, government). These identities can be partly stored as a digital twin (e.g. by recording biometric information plus some identifier/number, and then possibly linked to other information credentials or entitlements – e.g. citizenship, age, health, finance, educational records and so on).
These digital ecosystems can be designed to allow us to control (access to) such data, or they can be part of state and commercial surveillance. The trustworthiness of such ecosystems is highly questionable. I’ll walk through alternative designs and give examples of benefits and disadvantages, including threats (fake id, denial of service etc).
In this talk, I’ll also outline challenges, including future problems like the mutability of allegedly unique and persistent biometrics like iris or even DNA , and speculate about the possibility of reflecting social structures properly in designs to create more fair and resilient systems that might be more acceptable than many deployed or proposed today.
- Speaker: Professor Jon Crowcroft. Computer Lab, at the University of Cambridge
- Monday 02 March 2026, 18:00-19:00
- Venue: Bristol-Myers Squibb Lecture Theatre, Department of Chemistry.
- Series: Cambridge Philosophical Society; organiser: Beverley Larner.
Thu 21 Aug 14:30: GNN-ACLP: Graph Neural Networks Based Analog Circuit Link Prediction
Circuit link prediction identifying missing component connections from incomplete netlists is crucial in analog circuit design automation. However, existing methods face three main challenges: 1) Insufficient use of topological patterns in circuit graphs reduces prediction accuracy; 2) Data scarcity due to the complexity of annotations hinders model generalization; 3) Limited adaptability to various netlist formats. We propose GNN -ACLP, a graph neural networks (GNNs) based method featuring three innovations to tackle these challenges. First, we introduce the SEAL (learning from Subgraphs, Embeddings, and Attributes for Link prediction) framework and achieve port-level accuracy in circuit link prediction. Second, we propose Netlist Babel Fish, a netlist format conversion tool leveraging retrieval-augmented generation (RAG) with a large language model (LLM) to improve the compatibility of netlist formats. Finally, we construct SpiceNetlist, a comprehensive dataset that contains 775 annotated circuits across 10 different component classes. Experiments demonstrate accuracy improvements of 16.08% on SpiceNetlist, 11.38% on Image2Net, and 16.01% on Masala-CHAI compared to the baseline in intra-dataset evaluation, while maintaining accuracy from 92.05% to 99.07% in cross-dataset evaluation, exhibiting robust feature transfer capabilities.
- Speaker: Guanyuan Peter Pan
- Thursday 21 August 2025, 14:30-16:00
- Venue: Zoom (details TBA).
- Series: Foundation AI; organiser: Pietro Lio.
Thu 21 Aug 14:30: GNN-ACLP: Graph Neural Networks Based Analog Circuit Link Prediction Looking for co-author(s) for new analog circuit automation project ;)
Circuit link prediction identifying missing component connections from incomplete netlists is crucial in analog circuit design automation. However, existing methods face three main challenges: 1) Insufficient use of topological patterns in circuit graphs reduces prediction accuracy; 2) Data scarcity due to the complexity of annotations hinders model generalization; 3) Limited adaptability to various netlist formats. We propose GNN -ACLP, a graph neural networks (GNNs) based method featuring three innovations to tackle these challenges. First, we introduce the SEAL (learning from Subgraphs, Embeddings, and Attributes for Link prediction) framework and achieve port-level accuracy in circuit link prediction. Second, we propose Netlist Babel Fish, a netlist format conversion tool leveraging retrieval-augmented generation (RAG) with a large language model (LLM) to improve the compatibility of netlist formats. Finally, we construct SpiceNetlist, a comprehensive dataset that contains 775 annotated circuits across 10 different component classes. Experiments demonstrate accuracy improvements of 16.08% on SpiceNetlist, 11.38% on Image2Net, and 16.01% on Masala-CHAI compared to the baseline in intra-dataset evaluation, while maintaining accuracy from 92.05% to 99.07% in cross-dataset evaluation, exhibiting robust feature transfer capabilities.
Looking for co-author(s) for new analog circuit automation project ;)
- Speaker: Guanyuan Peter Pan
- Thursday 21 August 2025, 14:30-16:00
- Venue: Zoom (details TBA).
- Series: Foundation AI; organiser: Pietro Lio.
Mon 13 Oct 18:00: Our Chiral Universe
The fundamental laws of physics look different when reflected in a mirror. This is the statement that the laws of physics have a handedness, what physicists call chirality. This is one of the most important facts that we know about the universe, a fact that, remarkably, goes a long way to fixing the mathematical structure of the laws of nature. I will explain how we know about this handedness, why it’s so important, and why there are still several chiral mysteries that remain unsolved.
- Speaker: Professor David Tong
- Monday 13 October 2025, 18:00-19:00
- Venue: Bristol-Myers Squibb Lecture Theatre, Department of Chemistry.
- Series: Cambridge Philosophical Society; organiser: Beverley Larner.
Wed 17 Sep 11:00: LMB Seminar - Ribosome Upside Down: Co-translational Protein Biogenesis at the Tunnel Exit
Our group aims to understand the cellular mechanisms underlying the production of functional proteins. This complex process requires an intricate interplay between the ribosome and a growing number of cellular factors that regulate translation and co-translational protein biogenesis. One of the key topics of our research, which combines structural, biochemical, and biophysical approaches, is the biogenesis of cytosolic and membrane proteins controlled by the nascent polypeptide–associated complex (NAC). I will present recent results that reveal how NAC coordinates the activity of a network of factors and enzymes at translating ribosomes to facilitate virtually all aspects of protein biogenesis, including nascent chain processing, modification, folding, and cellular localization such as targeting to the ER.
- Speaker: Nenad Ban - ETH Zürich
- Wednesday 17 September 2025, 11:00-12:00
- Venue: In person in the Max Perutz Lecture Theatre (CB2 0QH) .
- Series: MRC LMB Seminar Series; organiser: Scientific Meetings Co-ordinator.
Wed 17 Sep 11:00: LMB Seminar - Ribosome Upside Down: Co-translational Protein Biogenesis at the Tunnel Exit
Our group aims to understand the cellular mechanisms underlying the production of functional proteins. This complex process requires an intricate interplay between the ribosome and a growing number of cellular factors that regulate translation and co-translational protein biogenesis. One of the key topics of our research, which combines structural, biochemical, and biophysical approaches, is the biogenesis of cytosolic and membrane proteins controlled by the nascent polypeptide–associated complex (NAC). I will present recent results that reveal how NAC coordinates the activity of a network of factors and enzymes at translating ribosomes to facilitate virtually all aspects of protein biogenesis, including nascent chain processing, modification, folding, and cellular localization such as targeting to the ER.
- Speaker: Nenad Ban - ETH Zurich
- Wednesday 17 September 2025, 11:00-12:00
- Venue: In person in the Max Perutz Lecture Theatre (CB2 0QH) .
- Series: MRC LMB Seminar Series; organiser: Scientific Meetings Co-ordinator.
Thu 23 Oct 11:30: Modeling and simulation of salt caverns: from lab to field scale
Underground man-made salt caverns are a proven technology for energy storage, and their usage is expected to increase in the coming years, following the current efforts towards energy transition. Upscaling salt caverns (e.g., systems of caverns) also raises concerns about safety and cavern integrity, which requires a careful lifetime assessment of their mechanical stability. In this context, this presentation examines the mechanical behavior and failure mechanisms of salt rocks, as well as methods for identifying situations that could compromise cavern integrity. The importance of a multiscale approach, spanning from laboratory experiments to field-scale simulations, is also discussed. Without diving into the mathematical details, a physical intuition is provided on how to compose a constitutive model to capture the different deformation mechanisms in salt rocks. Finally, the impact of different constitutive model choices and calibrations is analyzed in the light of numerical simulations.
- Speaker: Herminio Tasinafo Honorio, Delft University of Technology
- Thursday 23 October 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.
Tue 07 Oct 14:00: Monotone Circuit Complexity of Matching
We show that the perfect matching function on $n$-vertex graphs requires monotone circuits of size $2}$. This improves on the $n{\Omega(\log n)}$ lower bound of Razborov (1985). Our proof uses the standard approximation method together with a new sunflower lemma for matchings.
- Speaker: Bruno Cavalar (Oxford)
- Tuesday 07 October 2025, 14:00-15:00
- Venue: Computer Laboratory, William Gates Building, Room SS03.
- Series: Algorithms and Complexity Seminar; organiser: Tom Gur.
Tue 04 Nov 14:00: TBD
TBD
- Speaker: Or Meir (Sheffield)
- Tuesday 04 November 2025, 14:00-15:00
- Venue: Computer Laboratory, William Gates Building, Room SS03.
- Series: Algorithms and Complexity Seminar; organiser: Tom Gur.
Fri 19 Sep 14:00: Human-AI Ecosystems for Daily Health and Well-being
As the intelligence of everyday smart devices continues to evolve, they can already monitor basic health behaviors such as physical activities and heart rates. The vision of an intelligent health monitoring and intervention pipeline seems to be within reach. How do we get there? In this talk, I will introduce a comprehensive pipeline that connects AI, end-users, and health experts. For end-users, I will introduce our work that bridges behavior science theory-driven intervention designs and generalizable behavior models. I will also introduce my efforts on passive sensing datasets, human-centered algorithms & large language models (LLMs), as well as a benchmark platform that drives the community toward more robust and deployable health systems for both end-users and experts.
Biography: Dr. Xuhai “Orson” Xu is an Assistant Professor at Columbia University’s Department of Biomedical Informatics and a Visiting Faculty at Google, where he leads research at the crossroads of human-computer interaction, applied AI, and health. His work develops deployable AI algorithms and intelligent interventions that harness everyday sensor data and health records to monitor and improve well-being, while his human-centered pipeline unites AI, clinicians, patients, and the broader community in a collaborative ecosystem for improved care. Dr. Xu’s work has been recognized through numerous awards and widespread media coverage for its groundbreaking contributions to digital health and human-computer interaction, including several Best Paper and Best Artifact awards at top-tier venues such as ACM CHI and IMWUT , the Innovation and Technology Award, and media posts such as the Washington Post, Scientific American, and ACM News.
- Speaker: Xuhai “Orson” Xu, Columbia University
- Friday 19 September 2025, 14:00-15:00
- Venue: Computer Lab, LT2 and Online.
- Series: Mobile and Wearable Health Seminar Series; organiser: Cecilia Mascolo.
Mon 01 Sep 11:00: LMB Seminar - Mechanisms of Translational Control: From Viral Hijacking to Codon-Dependent mRNA Surveillance - In person only
In eukaryotic cells, the translation of mRNA into protein is tightly regulated at multiple levels. In this talk, I will present two complementary studies that explore distinct, yet mechanistically intertwined, aspects of translational control. The first part focuses on how the Hepatitis C virus (HCV) internal ribosomal entry site (IRES) manipulates host translation machinery to drive cap-independent viral protein synthesis. Through a series of high-resolution cryo-EM structures, we reveal how the IRES restructures eukaryotic initiation factor 3 (eIF3), with its core subunits being displaced by tight interaction with the IRES while the non-core subunits remain positioned on the ribosome. Unexpectedly, the N-terminal domain of the eIF3c subunit interacts with the 60S ribosomal subunit during elongation, suggesting that eIF3 plays roles beyond initiation, potentially extending to elongation, termination, and ribosome recycling. In the second part, we uncover a novel function of the RNA helicase DHX29 in regulating mRNA stability based on codon optimality. Using genome-wide CRISPR screening, selective ribosome profiling, and cryo-EM, we demonstrate that DHX29 binds near the A-site entrance of translating ribosomes and senses the decoding efficiency of incoming aminoacyl-tRNAs. This ribosomal interaction allows DHX29 to recruit the GIGYF2 •4EHP complex, thereby linking slow translation caused by non-optimal codons to mRNA decay pathways. Together, these studies highlight how both viral elements and host RNA -binding proteins reshape the translational landscape—either to hijack the host machinery or to maintain transcriptome integrity—offering new structural and mechanistic insights into translational regulation.
- Speaker: Takuhiro Ito, RIKEN Center for Integrative Medical Sciences
- Monday 01 September 2025, 11:00-12:00
- Venue: In person in the Max Perutz Lecture Theatre (CB2 0QH) .
- Series: MRC LMB Seminar Series; organiser: Scientific Meetings Co-ordinator.