#StateSpaceModels

Want to learn how to fit and validate #statespacemodels ? A team of great co-authors (@vianey_lb @len_thom +) and I put this review together: The appendix is full of R code to help you fit models to your data.

Generative Temporal Models with Spatial Memory for Partially Observed Environments It was really cool working with Marco Fraccaro on this project during his internship @DeepMindAI. Great work Marco! #statespacemodels #inference #spatialmemory #vae

Excited to share our #CVPR2025 paper, "GG-SSMs: Graph-Generating State Space Models", to be presented as a Highlight Paper in Nashville this week! PDF: Code: While #StateSpaceModels (#SSMs) are extremely powerful for sequential data, the one-dimensional processing paradigm severely limits their ability to model non-local interactions in high-dimensional data. Even advanced models like #Mamba, #Vim, and #VMamba, though offering selective scanning, remain constrained by predetermined paths, often failing to efficiently capture the high-dimensional interactions of data dependencies. Our new Graph-Generating State Space Models (GG-SSMs) directly address this limitation. We introduce a novel framework that dynamically constructs graphs based on inherent feature relationships, adapting to the unique structure of the data itself! By leveraging Chazelle's Minimum Spanning Tree (MST) algorithm, known for its near-linear time complexity, GG-SSMs enable robust feature propagation and efficiently model complex, long-range dependencies without prohibitive computational costs. Our contributions are as follows: * Integration of dynamic graph structures directly into the SSM framework, capturing complex spatial and temporal relationships. * SOTA results across 11 diverse datasets, including ImageNet, optical flow, event-based eye tracking, and time series! - ImageNet: Sets a new benchmark with 84.9% top-1 accuracy, outperforming prior SSMs by 1%. - KITTI-15 Optical Flow: Achieves the lowest error rate ever reported at 2.77%. - Eye Tracking: Improves detection rates by up to 0.33% with fewer parameters on event-based datasets. - Time Series: Demonstrates superior forecasting accuracy across six real-world datasets. * MST-based construction ensures linear computational complexity, making GG-SSMs highly scalable and efficient for large-scale applications. Kudos to @NikolaZubic5 ! Reference: Nikola Zubić and Davide Scaramuzza, GG-SSMs: Graph-Generating State Space Models, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, 2025., PDF: Code: Highlight Presentation. @ERC_Research @uzh_ifi @UZH_en @UZH_Science #GraphNeuralNetworks #ComputerVision #MachineLearning #AIResearch #DeepLearning #OpticalFlow #TimeSeries

📢 ICLR 2026 Paper 📢 ICLR 2026 GMamba: Enabling true global perception in vision state space models (linear complexity). 🔗 Paper: 💻 Code: #ICLR2026 #Mamba #StateSpaceModels #ComputerVision

Excited to be in Vienna next week for @icmlconf, and ws on Next Gen. of Seq. Modeling Architectures, where will present 3 papers on converging #StateSpaceModels and #ReservoirComputing (yes, we did it💪). Looking forward to engaging discussions! 😍😍 #AI #MachineLearning

We are excited to share our #CVPR2024 paper "State Space Models for Event Cameras"! Code released! Traditional deep neural networks struggle with event-camera data at higher frequencies, leading to a trade-off between accuracy and speed. We introduce #StateSpaceModels (#SSMs) to event cameras. These enable the networks to adapt to varying frequencies. Our results show that SSM-based models train 33% faster than state-of-the-art methods while preserving task performance at higher frequencies at test time, thus resolving the problem of traditional #RNNs and #Transformers. We show that SSM-based models offer the fastest inference times and require the least number of parameters (see Table 1 in the paper). We believe this is a step towards more efficient and adaptable event-based neural networks in real-world applications. Kudos @NikolaZubic5 @MathiasGehrig ! Reference: Nikola Zubić, Mathias Gehrig, Davide Scaramuzza "State Space Models for Event Cameras" IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2024. Spotlight Presentation. Paper: Code: Video: #SequenceModeling #SSM #SSMs #RNN #RNNs #Transformers #NeuralNetworks #MachineLearning #AIResearch #Innovation #DeepLearning #eventcamera #eventcameras #neuromorphic #neuromorphichardware

Introducing 'HumMUSS: Human Motion Understanding using State Space Models'! Our new model offers adaptable frame rates, faster real-time predictions, and high accuracy in motion analysis. #ComputerVision #StateSpaceModels #CVPR2024

Now online in @ESAMonographs: A guide to state–space modeling of ecological time series With #OpenData in @datadryad #StateSpaceModels #EcologicalTimeSeries #StatisticalEcology @AugerMethe @vianey_lb @len_thom @JoannaMFlemming @DinCole2017 @ubcscience

Meet us at #CVPR2024 this week! We will present several papers on #Nerf deblurring, #StateSpaceModels, and Motion Estimation on Manifolds for #eventcameras at the main conference and workshops! Full list with times, rooms, and links to PDFs, Code, and Videos: @marcocannici @DanielGehrig6 @NikolaZubic5

Exploring the latest in state-space models from ICLR, I've noted key future tasks. Impressed by Mamba & FlashFFTConv's advancements in scaling & speed. What's the next big leap in this field? #Mamba #StateSpaceModels #ICLR

Great few days at #ETNsymposium2024 listening to lots of #inspiring presentations and sharing our work on the #patter #rstats #JuliaLang package for modelling animal-tracking data! @EawagResearch #trackingnotslacking #AnimalMovement #SSM #SMC #ParticleFilter #ecology 1/6

Edward Lavender introduces us to the “patter” package for acoustic telemetry 🤓🐟 #ETNsymposium2024

#patter fits #StateSpaceModels to #AnimalTracking data using #ParticleFiltering and #Smoothing algorithms, generating refined #maps of #SpaceUse & #residency estimates & supporting downstream #analyses in #AcousticTelemetry systems. 2/6

Generalising #ML models for building energy forecasting relies on representation of the governing dynamics in the data. This paper introduces a transfer learning approach that exploits the geometry of Physics-based #StateSpaceModels in a subspace-based domain adaptation framework

A physics-based domain adaptation framework for modeling & forecasting building #EnergySystems @ZackXConti, @RuchiChoudhary_ & Luca Magri → #EnergyForecasting #DomainAdaptation #DynamicalSystems #EnergyEfficiency #ML #Buildings #TransferLearning

The third part of my parameter estimation on state space models guide is here: making a proposal distribution for state space model parameter! #research #statespacemodels #julialang #mcmc #pmcmc

I am planning to highlight some of my research projects and related topics as blog posts over the next weeks. The series begins with an introduction to state space models — click below to take a look! #research #statespacemodels #HMM #julialang

Séminaire annuel du groupe Dynamique de populations de l'@oncfs animé par @GSouchay et @CCalenge sur les analyses de tendances. #rstat #statespacemodels #poptrend

RESEARCH to look forward to: @AugerMethe and team of "great co-authors" have put a review of #statespacemodels. She says it has lots of fisheries models in it, a whale model, and the image is the track of a polar bear. Preprint only at moment

Super excited to join the South Africa Biodiversity Data Pipeline for Wetlands and Waterbirds (BIRDIE) project, after yet another bumpy year. #StateSpaceModels, #OccupancyModels, #CitizentScience, #rstats, #DataPipeline. The project ticks lots of boxes!

I created a short series of four blog posts, explaining parameter inference on state space models. Here is the first article, an overview on different estimation paradigms: #research #statespacemodels #pmcmc

As State Space Model breakthroughs arrive and grow and get us back to efficient modeling of hidden states, transformers will be seen as 'fake it till you make it' placeholders for RNN. With gratitude, though. @_albertgu #statespacemodels

The second part of my parameter estimation on state space models guide is here: particle filtering for hidden Markov models! #research #statespacemodels #julialang #particlefiltering #pmcmc