News

• December 2024: We presented our examinations of low cost approaches to generate new realizations of large-scale climate ensembles with conditional variational autoencoders at the American Geophysical Union (AGU) Fall Meeting.
   
• December 2023: We presented our work exploring the decomposition of variability of Antarctic sub-shelf melt via generalized clustering with kernel embeddings at the American Geophysical Union (AGU) Fall Meeting.
   
• December 2022: We presented our first results generating additional data in a data scarce domain, namely sub-shelf ice melt in Antarctica using recurrent GANs on spatiotemporally clustered melt data (E3SM model output) at the American Geophysical Union (AGU) Fall Meeting in December! Thanks for such positive reception and encouraging feedback!
   
• July 2022: I was honored to be invited to the SIAM Conference on the Mathematics of Planet Earth, symposium on Statistical Models and Learning Methods in Wildfire Science to give a talk on Probabilistic Machine Learning, Bayesian Inference, and Remote Sensing for Environmental Data. (SIAM video).
   
• May 2022: Our work towards generating stationary time-series data of Antarctic Ice Sheet ice shelf melt from limited climate model output was accepted to the Climate Informatics Conference! This is ongoing work in an interdisciplinary collaboration with polar climate experts, sparked at the Machine Learning and the Physics of Climate program at KITP.
   
• November-December 2021: I was invited to participate as a core member in the Machine Learning and the Physics of Climate (and give a talk, see below) at the University of California Santa Barbara (UCSB) within the Kavli Institute for Theoretical Physics (KITP).
   
• November 2021: I was invited and honored to give a talk at the first Machine Learning for Climate Conference hosted by the UCSB within the KITP. I talked about deep learning and fine segmentation of dead trees within forests, which is critical to understanding carbon sequestration and wood decay, effects on biodiversity, the climate, and much more. This work corresponds to this paper at the NeurIPS workshop on Climate Change.
   
• December 2021: We will be presenting our paper entitled A hybrid convolutional neural network/active contour approach to segmenting dead trees in aerial imagery at the NeurIPS workshop on Tackling Climate Change with Machine Learning. I also gave a more comprehensive talk at this conference. Come check it out!
   
• July 2021: Our work exploring instance segmentation of fallen trees with satellite imagery and the respective contribution of carbon stock has been published in ISPRS Journal of Photogrammetry and Remote Sensing!
   
• October 2020: Our work exploring instance segmentation of standing trees using GAN-based shape priors was published at ISPRS Conference on Photogrammetry and Remote Sensing!
   
• September 2020: Our paper using U-Nets for learning and inference of dense representation of multiple air pollutants from satellite imagery was accepted to the Climate Informatics Conference 2020 and selected to be a highlight Talk!
   
• July 2020: Our paper investigating Covid-19 lethality / morbidity based on pollution as a transmission vector using readily available satellite imagery -- In the Danger Zone: U-Net Driven Quantile Regression can Predict High-risk SARS-CoV-2 Regions via Pollutant Particulate Matter and Satellite Imagery -- was accepted to the ICML Workshop on Healthcare Systems, Population Health, and the Role of Health-Tech! : I joined the AI Geospatial Research group, "TAIGR", at TomTom in Berlin lead by Soeren Sonnenburg.
   
• May 2018: I received my Dr. rer. nat. (Doctor of Natural Sciences) in Computer Science (Machine Learning) from the Technische Universität Berlin (Technical University of Berlin) with the thesis entitled Large-scale Approximate EM-style Learning and Inference in Generative Graphical Models for Sparse Coding!
   
• August 2017: Our work proposing a method for efficient inference in genererative models with high dimensional latent variables entitled GP-select: Accelerating EM using adaptive subspace preselection (arXiv) made the front cover of and was published in Neural Computation! -->
   
   

Publications  [Google scholar]

Articles:

• Polewski, P., Shelton, J., Yao, W., and Heurich, M.: Instance segmentation of fallen trees in aerial color infrared imagery using active multi-contour evolution with fully convolutional network-based intensity priors.
  ISPRS Journal of Photogrammetry and Remote Sensing, (178): 297--313, (2021).
   
• Shelton, J., Gasthaus, J., Dai, Z., Luecke, J., and Gretton, A.: GP-select: Accelerating EM using adaptive subspace preselection.
  Neural Computation, 29 (8): 2177--2202, (2017).
  [PDF - old arXiv pdf]
   
• Shelton, J. A., A.-S. Sheikh, J. Bornschein, Sterne, P. and J. Lücke: Nonlinear Spike-and-Slab Sparse Coding for Interpretable Image Encoding.
  PLOS ONE, May, (2015).
   
• Sheikh, A-S., Shelton, J. A., Luecke, J: A Truncated EM Approach for Spike-and-Slab Sparse Coding.
  Journal of Machine Learning (JMLR), 15(Aug): 2653--2687, (2014).
   
  
• M. B. Blaschko, Shelton, J. A., Bartels, A., Lampert, C.H., and A. Gretton: Semi-supervised Kernel Canonical Correlation Analysis with Application to Human fMRI.
  PRL, 32(11): 1572--1583, (2011).
   

Conference papers:

• Shelton, J., Polewski, P., Yao, W., and Heurich, M.: A hybrid convolutional neural network/active contour approach to segmenting dead trees in aerial imagery.
  NeurIPS workshop on Tackling Climate Change with Machine Learning, Virtual Worldwide, (2021).
  Invited talk at Machine Learning for Climate Conference 2021.
  [talk - slides - paper]
   
• Shelton, J., Polewski, P., Yao, W.: In the Danger Zone: U-Net Driven Quantile Regression can Predict High-risk SARS-CoV-2 Regions via Pollutant Particulate Matter and Satellite Imagery.
  ICML Workshop on Healthcare Systems, Population Health, and the Role of Health-Tech, Virtual Worldwide, (2020).
   
• Shelton, J., Polewski, P., Yao, W.: U-Net For Learning And Inference Of Dense Representation Of Multiple Air Pollutants From Satellite Imagery.
  Climate Informatics, Virtual Worldwide, (2020).
  Highlight Talk.
  [paper - slides]
   
• Polewski, P., Shelton, J., Yao, W., and Heurich, M.: Segmentation of single standing dead trees in high-resolution aerial imagery with generative adversial network-based shape priors.
  International Arch. Photogramm. Remote Sensing Spatial Inf. Sci., XLIII-B2-2020: 717--723, (2020).
   
• Shelton, J.A., Sterne, P., J. Bornschein, A.-S. Sheikh, and J. Lücke: Why MCA? Nonlinear sparse coding with spike-and-slab prior for neurally plausible image encoding.
  Proceedings of the Twenty-Sixth Annual Conference on Neural Information Processing Systems, (NIPS 2012).
  [paper - poster]
   
• Shelton, J.A., J. Bornschein, A.-S. Sheikh, P. Berkes, and J. Lücke: Select and Sample - A Model of Efficient Neural Inference and Learning.
  Proceedings of the Twenty-Fifth Annual Conference on Neural Information Processing Systems, (NIPS 2011).
  [paper - slides - poster]
   
• Dai, Z., Shelton, J., Bornschein, J., Sheikh, A. S., and Lücke, J: Combining approximate inference methods for efficient learning on large computer clusters.
  NIPS workshop on Big Learning: Algorithms, Systems, and Tools for Learning at Scale, (2011).
   
• Blaschko, M. B., J. A. Shelton and A. Bartels: Augmenting Feature-driven fMRI Analyses: Semi-supervised Learning and Resting State Activity.
  Proceedings of the Twenty-Third Annual Conference on Neural Information Processing Systems (NIPS 2009).
  [paper - slides - poster]
   

Technical reports:

• Shelton, J.A., Gasthaus, J., Dai, Z., Lücke, J., and Gretton, A.: GP-select: Accelerating EM using adaptive subspace preselection
  arXiv:1412.3411 [stat.ML], (2014).
   
• A.-S. Sheikh, Shelton, J.A., and J. Lücke: A Truncated Variational EM Approach for Spike-and-Slab Sparse Coding.
  arXiv:0594372, (2012).
   
• Shelton, J., M. B. Blaschko and A. Bartels: Semi-supervised Subspace Analysis of Human Functional Magnetic Resonance Imaging Data.
  Max Planck Institute for Biological Cybernetics Tech Report (185) (2009)
   

Theses:

• Doctoral Thesis: Large-scale Approximate EM-style Learning and Inference in Generative Graphical Models for Sparse Coding.
  Dr. rer. nat. Technische Universität Berlin, Dept. of Software Engineering and Theoretical Computer Science, Group of Klaus-Robert Mueller, Intelligent Data Analysis. (05 2018)
   
• Masters Thesis: Semi-supervised Subspace Learning and Application to Human Functional Magnetic Brain Resonance Imaging Data.
  M.S. Thesis, Universität Tübingen, Dept. of Computer Science, and Max Planck Institute for Biological Cybernetics, Dept. Schölkopf. (07 2010)
   

Posters:

• Shelton, J. A., Robel, A. A., Hoffman, M., and Price, S.: Generating new realizations of large-scale climate ensembles with conditional variational autoencoders.
  American Geophysical Union (AGU), Fall Meeting, (12 2024).
  [poster]
   
• Shelton, J. A., Robel, A. A., Hoffman, M., and Price, S.: Decomposing Antarctic Sub-shelf Melt Variability using Generalized Clustering with Kernel Embeddings.
  American Geophysical Union (AGU), Fall Meeting, (12 2023).
  [poster]
   
• Shelton, J. A., Robel, A. A., Hoffman, M., and Price, S.: Generating Antarctic sub-shelf melt using recurrent neural network-based Generative Adversarial Models on pixel clusters.
  American Geophysical Union (AGU), Fall Meeting, (12 2022).
  [abstract - poster]
   
• Shelton, J. A., Robel, A. A., Hoffman, M., and Price, S.: Generating Antarctic sub-shelf melt using recurrent neural network-based Generative Adversarial Models on pixel clusters.
  Machine Learning for Polar Regions Workshop, 2022.
   
• Shelton, J. A., Robel, A. A., Hoffman, M., and Price, S.: Towards generating stationary realizations of simulated Antarctic ice shelf melt rates from limited model output.
  Climate Informatics, 2022.
   
• Shelton, J., Gasthaus, J., Dai, Z., Lücke, J., and Gretton, A.: GP-select: Accelerating EM using adaptive subspace preselection.
  Women in Machine Learning Workshop (WiML 2014), in conjunction with (NIPS 2011), Montreal, Quebec, (2014).
   
• Shelton, J., and Christoph Lampert: Approximate Inference with δ-insensitive Marginal Loss.
  Women in Machine Learning Workshop (WiML 2013), in conjunction with (NIPS 2013), Lake Tahoe, Nevada (12 2013).
   
• Lücke, J., Shelton, J., Bornschein, J., Sterne, P., Berkes, P., and Sheikh, A-S: Combining Feed-Forward Processing and Sampling for Neurally Plausible Encoding Models.
  Cosyne 2013, ID: III-28. (2013).
   
• Shelton, J. A., Sterne, P., J. Bornschein, A.-S. Sheikh, and J. Lücke: Why MCA? Nonlinear sparse coding with spike-and-slab prior for neurally plausible image encoding.
  Women in Machine Learning Workshop (WiML 2012), in conjunction with (NIPS 2012), Lake Tahoe, Nevada, (2012).
   
• Dai, Z., Shelton, J., Bornschein, J., Sheikh, A. S., and Lücke, J: Combining approximate inference methods for efficient learning on large computer clusters.
  NIPS workshop on Big Learning: Algorithms, Systems, and Tools for Learning at Scale, (2011).
   
• Bornschein, J., Shelton, J. A., Sheikh, A. S., and Lücke, J: The Maximal Causes of Binary Data.
  Bernstein Conference on Computational Neuroscience (BCCN), (2011).
   
• Shelton, J.A., J. Bornschein, A.-S. Sheikh, P. Berkes, and J. Lücke: Select and Sample - A Model of Efficient Neural Inference and Learning. Women in Machine Learning Workshop (WiML 2011), in conjunction with (NIPS 2011), Malaga, Spain, (2011).
   
• Shelton, J. A., M. B. Blaschko, A. Gretton, J. Müller, E. Fischer and A. Bartels: Similarities in Resting State and Feature-driven Activity: Non-parametric Evaluation of Human fMRI.
  NIPS Workshop on Learning and Planning from Batch Time Series Data, (2010).
   
• Shelton, J.A., Blaschko, M. B., and A. Bartels: Augmentation of fMRI Data Analysis using Resting State Activity and Semi-supervised Canonical Correlation Analysis.
  Women in Machine Learning Workshop (WiML 2010), in conjunction with (NIPS 2010), Vancouver, BC, Canada (2010). Highlight Talk.
   
• Shelton, J. A., M. B. Blaschko, C. H. Lampert and A. Bartels: Semi-supervised Analysis of Human fMRI.
  Berlin Brain-Computer Interface Workshop (BBCI), (2009).
   

Talks

• Shelton, J. A.: Probabilistic machine learning for uncertainty representation and applications to neural coding in biological sensory systems.
  International Conference on Mathematical Modeling and Analysis of Populations in Biological Systems (10 2022).
  University of Lousiana, Lafayette, USA. Invited talk.
   
• Shelton, J. A.: Probabilistic Machine Learning, Bayesian Inference, and Remote Sensing for Environmental Data. (SIAM video).
  Statistical Models and Learning Methods in Wildfire Science (7 2022).
  (SIAM) Conference on Mathematics of Planet Earth, Pittsburgh, USA. Invited talk.
   
• Shelton, J. A.: Deep learning and energy models for fine dead wood segmentation.
  Machine Learning for Climate Conference (11 2021).
  University of California, Santa Barbara, Kavli Institute for Theoretical Physics, California, USA. Invited talk.
   
• Shelton, J. A.: U-Net For Learning And Inference Of Dense Representation Of Multiple Air Pollutants From Satellite Imagery.
  Climate Informatics 2020, (Worldwide), (9 2020). Highlight talk.
   
• Shelton, J. A.: Lecture series on Bayesian Reasoning and Probabilistic Modelling. Invited lecturer, (2014-2015).
  Data Science Retreat, Berlin, Germany. Twice invited lecturer.
   
• Shelton, J. A.: Select and Sample.
  Technical University Darmstadt, Darmstadt, Germany, (6 2012). Invited talk.
   
• Shelton, J. A.: Select and Sample.
  Institute for Science and Technology (IST) Austria, Vienna, Austria, (2 2012). Invited talk.
   
• Shelton, J. A.: Select and Sample.
  Radbound University Nijmegen, Nijmegen, The Netherlands, (1 2012). Invited talk.
   
• Shelton, J. A.: Semi-supervised Kernel Canonical Correlation Analysis of Human Functional Magnetic Resonance Imaging Data.
  Women in Machine Learning Workshop (WiML 2009), Vancouver, BC, Canada (12 2009). Invited talk.
   

Participation

• Judge:
  Outstanding Student Presentation Award (OSPA)
  American Geophysical Union (AGU) Fall Meeting 2023
   
• Chair:
  European Conference on Machine Learning (ECML) 2021 session on Generative Models.
   
• Reviewer:
  Neural Information Processing Systems (NeurIPS), Artificial Intelligence and Statistics (AISTATS), , IEEE Transactions on Pattern Analysis and Machine Intelligence (IEEE-TPAMI), IEEE Transactions on Knowledge and Data Engineering (IEEE-TKDE), Environmental Data Science, Climate Informatics
   
• Teaching:
  Data Science retreat, 2014-2015, Berlin, Germany
  Lecture series on Bayesian Reasoning and Probabilistic Modelling.
   

Code

• GP-select Demo on Gaussian mixture models.
  An illustration of the approximate inference method corresponding to our Neural Computation paper, GP-select: Accelerating EM using adaptive subspace preselection.