Deep Learning workshops

This workshop examines practical implementations and evaluation of autonomous machine intelligence in data mining. Themes include predictive world models, joint-embedding predictive architectures, self-supervised multimodal learning, intrinsic motivation and scalable analytics systems. It also considers interpretability, robustness and responsible deployment. The programme combines keynotes, research presentations and a discussion of the gap between architectural ideas and real-world systems. Workshop W05 runs from 08:30 to 12:00 local time. Paper submissions closed on 20 August and camera-ready papers were due on 2 October 2026.

A two-day live course introduces fully connected neural networks, convolutional networks for computer vision, recurrent networks for text and speech, transfer learning and large language models. Participants use PyTorch to build and train models on high-end GPUs through an HPC cluster, connecting from their own computers to the Austrian Scientific Computing JupyterHub. Temporary cluster accounts support the exercises, and a short pre-assignment prepares participants before the course. The English-language course is intended for software developers, data scientists and researchers moving from using AI tools to building and adapting models. Basic Python programming and Linux command-line skills are required; previous deep-learning knowledge is not. Sessions run 09:00–16:00 CET on 23 and 24 November 2026. Participation is free for qualifying attendees from academia, industry or public administration in EU or EuroHPC JU member countries. Institutional affiliation is verified at registration, and a phone number is needed for cluster two-factor authentication.

Build and evaluate deep-learning models for biological prediction using Python and Jupyter notebooks. Lectures and collaborative exercises develop convolutional networks for classification, regression and image segmentation, with attention to preprocessing, augmentation, class imbalance, cross-validation and overfitting. Later sessions explore learning-rate and training-duration choices, recurrent networks and applications to omics data. Participants interpret model performance and discuss their own research problems. 2026-11-23–2026-11-27, 14:00–20:00 daily, Europe/Berlin (CET, UTC+01:00). Online course; fee €530. A biology background is expected; basic Python and Linux familiarity is helpful but not required.

Apply machine learning throughout drug discovery, from biological target identification to molecular screening. The workshop explains model design, training, evaluation and prediction interpretation, then develops deep-learning and graph-neural-network methods in PyTorch. Practical work turns molecular and omics data into model inputs, implements a graph model from a scientific paper, and examines disease classification and target discovery. The final day builds pipelines from research databases and introduces language-model fine-tuning using biological data. Live online, 30 November–3 December, 14:00–19:00 Europe/Berlin (CET). Fee €480. Aimed at MSc/PhD students and early-career scientists, with basic Python, molecular biology, chemistry and data-analysis knowledge. Previous machine-learning experience is unnecessary.

This workshop connects genomic discovery with prediction of epilepsy outcomes. It examines genome-based multi-omics in multigenerational cohorts, multimodal deep learning that combines clinical and genomic information to predict antiseizure medication response, and polygenic risk scores integrated with machine learning to estimate disease severity and seizure control. The presentations link these complementary approaches to the problem of clinical heterogeneity and assess their practical role in more individualized treatment. Sunday 6 December, 14:00–15:30 Mountain Standard Time (America/Denver). Part of the 2026 American Epilepsy Society Annual Meeting. In-person meeting registration and a free myAES account are required. Standard full-meeting rates are USD 800 for member scientists and USD 1,075 for non-member scientists; student rates are USD 250 and USD 360, respectively. Other professional and one-day categories are available. Seating priority goes to registered attendees.

NeurReps, Symmetry and Geometry in Neural Representations, is a NeurIPS workshop that brings together mathematicians, machine learning researchers and neuroscientists working to uncover the geometric principles shared by brains and machines. In neuroscience, mounting evidence suggests that neural circuits encode task and environmental structure through low-dimensional manifolds, conserved symmetries and structured transformations; in deep learning, sparsity, equivariance and compositionality are guiding the development of more generalisable and interpretable models. Following editions from 2022 to 2025, NeurReps 2026 expands into geometric representation steering and geometric mechanistic interpretability and introduces a Findings track to foster collaboration between experimentalists and theorists, alongside an archival proceedings track published in a dedicated PMLR volume. The programme combines invited talks, contributed papers and a panel session on topology, geometry and neuroscience towards a unified framework. The workshop runs from 8:00 to 17:00 AEDT on Friday 11 December 2026 in Hall 4 of the International Convention Centre (ICC) Sydney. It is organised by Francisco Acosta, David Klindt, Louisa Cornelis, Alison Pouplin, Nina Miolane and Fatih Dinc.

UniReps asks when, how and why different neural models learn the same representations. Findings in neuroscience and artificial intelligence show that biological brains and artificial models often form similar representations when exposed to similar stimuli. The workshop brings machine learning, neuroscience and cognitive science together to discuss theoretical findings, empirical evidence and practical applications of this phenomenon: learning dynamics in neuroscience, identifiability in the functional and parameter spaces of artificial networks, representation alignment for model merging, stitching and reuse and for multimodal settings, and the invariances that emerge across learning processes, with the aim of unifying internal representations across distinct neural models. The 2026 workshop takes place in Paris on 12 December 2026, coinciding with NeurIPS in Paris. Invited speakers are Marco Baroni (Pompeu Fabra University), Alex Cayco-Gajic (École Normale Supérieure), Sophia Koepke (TUM/University of Tübingen), Joao Sacramento (Google) and Greta Tuckute (Harvard University). Technical paper submissions were due on 10 October 2026 (AoE).

Foundation Models for the Brain and Body is the second edition of a NeurIPS workshop on large-scale pretrained AI models of neural and physiological signals. Advances in brain interfacing and wearable technologies, including EEG, intracortical electrophysiology, EMG, MEG and ECG, now yield large and diverse biosignal datasets collected across real-world contexts, but these timeseries are noisy, heterogeneous and recorded under variable conditions across subjects, devices and environments. The workshop brings together neuroscientists, biomedical engineers, wearable-technology researchers and machine learning experts to develop foundation models that generalise across applications from brain-computer interfacing to health monitoring. This edition broadens the scope to embodied intelligence, with foundation models for movement, motor control and closed-loop interaction between brain, body and machine, and adds a panel discussion on benchmarking that compares evaluation practices across modalities. Invited speakers and panelists are Giulia Lioi (IMT Atlantique), Simon Dahan (Meta), Chethan Pandarinath (Emory, Georgia Tech), A. Ali Heydari (Google), Hubert J. Banville (Meta), Konstantinos Kontras (KU Leuven, MIT) and Sergey D. Stavisky (UC Davis). The workshop runs from 8:00 to 17:00 AEDT on Saturday 12 December 2026 in Hall 1 at NeurIPS 2026 in Sydney. Papers and travel-award applications were due on 5 September 2026 and demos on 19 September 2026.

Mechanistic cancer models and data-driven learning offer complementary ways to understand tumour behaviour and improve individualized treatment. This workshop explores how mathematical modelling can be combined with machine and deep learning, bringing together data scientists and mathematical oncologists. Plenary lectures survey current developments, while collaborative group projects develop algorithms that integrate mechanistic knowledge with observations. The programme aims to turn methodological exchange into sustained collaborations and more informative approaches to personalized cancer care.

Apply neural networks to biological sequence analysis, including sequence classification, functional-feature discovery, source tracking and sequence generation. Hands-on notebooks introduce sequence encoding and language-model concepts, progressing from convolutional networks and LSTMs to attention and Transformers. Exercises use TensorFlow and Keras to connect model training and performance evaluation with the biological interpretation of genomic data. 2027-01-25–2027-01-27, 14:00–20:00 daily, Europe/Berlin (CET, UTC+01:00). Online course; fee €480. Participants should be familiar with Python or R and basic biology or genomics.

A practical doctoral course in rational, physics-based and AI approaches to protein design. Topics include coiled-coil design, Rosetta, structure-based generative models, protein language models, screening and selection, proteins for extreme or cellular environments, molecular recognition, dynamic complexes, switches and motor proteins. Participants construct algorithms and use existing design software through readings, problem-based exercises and discussion. The programme also addresses translation, innovation and responsible research. In-person teaching runs 1–12 February 2027 in Copenhagen; the published course span is 09:00 on the first day to 17:00 on the last. Teaching is in English and awards 8.1 ECTS. Prior protein-chemistry training, Python, Unix and a laptop are required. The fee is DKK 16,080; eligible Danish-university and NorDoc PhD students are exempt. Danish PhD students have seat priority; other applicants wait until the 3 January deadline before remaining places are allocated.

Researchers learn the assumptions and implementation principles behind machine-learning models used in cancer genomics. The five-day programme develops hidden Markov approaches for copy-number inference, mixture models for tumour subclones, hierarchical models for mutation signals and autoencoders for single-cell data. Lectures alternate with discussion, coding practice and a final project, emphasizing inference of latent biological states from noisy or incomplete measurements. Participants may discuss suitable datasets they own. Online teaching runs 14:00–19:00, Monday through Friday.

This twelve-day workshop develops AI and machine-learning methods for data-intensive astronomy and cosmology. It progresses from foundational instruction through applications in galaxy formation, dark energy, time-domain observations and cosmological simulations to collaborative research planning. Deep learning, generative modelling, uncertainty estimation and workflow automation are considered alongside interpretability and disciplinary training gaps. The programme targets PhD students, postdocs and junior faculty in astronomy or AI-related fields; exceptional undergraduate and master’s applicants may be considered. Python, statistics and machine-learning experience are helpful for the practical work. ICTS provides accommodation for outstation participants.

Develop practical Python skills for collecting, processing and visualising biological data, then train and evaluate machine-learning models using public datasets. Teaching covers APIs, reproducible preprocessing, statistical analysis, neural networks, transformers and potential uses of large language models in life science. Mentored small-group projects apply the methods to biological questions. The course particularly suits PhD students and early-career researchers with beginner Python; previous machine-learning experience is unnecessary. There are 30 selected places, costing £975 for academia or £1,275 for industry. Applications open 30 November 2026 and close at 23:59 UK time on 28 February 2027, with decisions by 15 March.

This hands-on course trains microscopy image-analysis practitioners to apply deep learning to their own data and research problems. It addresses the gap between successful neural-network methods and the practical difficulties of training and deploying them. Applications span image restoration, classification and segmentation, object detection and morphological characterization, high-throughput imaging, and large-scale three-dimensional processing. The course focuses on practical use of these methods for microscopy rather than treating model development as a purely theoretical subject.

The NeuroData AI Summer School, formerly NeuroDataReHack, is a one-week residential school on reanalysing open neurophysiology data and applying AI methods to neural data, held from 19 to 24 July 2027 at HHMI's Janelia Research Campus in Ashburn, Virginia. Foundations sessions cover how NWB files are organised, how to find and stream data from the DANDI Archive, and analysis tools built on the standard such as SpikeInterface, Pynapple and Neurosift, using large public datasets. AI-methods sessions cover foundation models for neural activity, coding with large language models and AI agents, decoding and dimensionality reduction, and how to check whether a method developed on one lab's recordings works on another lab's data. Mornings are lectures and afternoons are team projects formed around shared datasets, and lectures are recorded and published. The school is intended for PhD students, postdoctoral researchers and principal investigators with basic Python or MATLAB programming and neurophysiology research experience; preference goes to applicants with a specific project reanalysing existing NWB data, and past NeuroDataReHack participants are not eligible. It is in person only for about 35 participants, with no registration or application fee, on-campus lodging and meals provided, and up to $550 of travel costs reimbursed. Applications close on 29 January 2027 and are reviewed on a rolling basis, with every applicant hearing back by the end of February 2027. The school is supported by NINDS award R25NS149357.

Biological mechanisms often have to be reconstructed from indirect measurements such as microscopy images, gene activity and cellular time series. This workshop examines inverse problems in living systems, where noise, complexity and diversity challenge existing mathematical approaches. Mathematicians, physicists, computer scientists and biologists will explore methods including deep learning and models informed by physical constraints to infer hidden cellular, genetic and neural mechanisms from experimental data.