{"id":345,"date":"2013-02-26T15:15:36","date_gmt":"2013-02-26T15:15:36","guid":{"rendered":"http:\/\/yingzhenli.net\/home\/en\/?page_id=345"},"modified":"2025-09-25T14:31:03","modified_gmt":"2025-09-25T14:31:03","slug":"publications","status":"publish","type":"page","link":"https:\/\/yingzhenli.net\/home\/en\/?page_id=345","title":{"rendered":"Publications"},"content":{"rendered":"

Working papers<\/strong><\/span><\/p>\n

Wuhao Chen, Zijing Ou<\/a> and Yingzhen Li<\/strong>. Neural Flow Samplers with Shortcut Models<\/a>. In submission. 2025.<\/p>\n

Harrison Zhu<\/a>, Adam Howes<\/a>, Owen van Eer, Maxime Richard, Yingzhen Li<\/strong>, Dino Sejdinovic<\/a> and Seth Flaxman<\/a>. Aggregated Gaussian Processes with Multiresolution Earth Observation Covariates<\/a>. 2022.<\/p>\n

Wenbo Gong<\/a> and Yingzhen Li<\/strong>.\u00a0Interpreting Diffusion Score Matching using Normalizing Flow<\/a>. ICML 2021 INNF+ workshop.<\/span><\/p>\n

Refereed conference papers<\/strong><\/span><\/p>\n

\n
\n
\n

Zijing Ou<\/a>, Ruixiang Zhang and Yingzhen Li<\/strong>. Discrete Neural Flow Samplers with Locally Equivariant Transformer<\/a>. Neural Information Processing Systems (NeurIPS), 2025.<\/p>\n

I. Shavindra Jayasekera, Jacob Si, Filippo Valdettaro, Wenlong Chen<\/a>, Aldo A. Faisal and Yingzhen Li<\/strong>. Variational Uncertainty Decomposition for In-Context Learning<\/a>. Neural Information Processing Systems (NeurIPS), 2025.<\/p>\n

Yohan Jung, Hyungi Lee, Wenlong Chen<\/a>, Thomas M\u00f6<\/span>llenhoff, Yingzhen Li<\/strong>, Juho Lee, Mohammad Emtiyaz Khan<\/a>. Compact Memory for Continual Logistic Regression. Neural Information Processing Systems (NeurIPS), 2025.<\/p>\n

Naoki Kiyohara, Edward Johns<\/a> and Yingzhen Li<\/strong>. Neural Stochastic Flows: Solver-Free Modelling and Inference for SDE Solutions. Neural Information Processing Systems (NeurIPS), 2025.<\/p>\n

Wenlong Chen<\/a>*, Naoki Kiyohara*, Harrison Bo Hua Zhu<\/a>*, Jacob Curran-Sebastian, Samir Bhatt and Yingzhen Li<\/strong>. Recurrent Memory for Online Interdomain Gaussian Processes<\/a>. Neural Information Processing Systems (NeurIPS), 2025.<\/p>\n

Xinzhe Luo<\/a>, Yingzhen Li<\/strong> and Chen Qin<\/a>. Unsupervised Accelerated MRI Reconstruction via Ground-Truth-Free Flow Matching<\/a>. Information Processing in Medical Imaging (IPMI), 2025. (oral)<\/p>\n

Manduchi et al. On the Challenges and Opportunities in Generative AI<\/a>. Transactions on Machine Learning Research (TMLR), 2025.<\/p>\n

Carles Balsells Rodas<\/a>, Xavier Sumba, Tanmayee Narendra, Ruibo Tu, Gabriele Schweikert, Hedvig Kjellstrom<\/a> and Yingzhen Li<\/strong>. Causal Discovery from Conditionally Stationary Time Series<\/a>. International Conference on Machine Learning (ICML), 2025.<\/p>\n

Wenlong Chen<\/a>, Bolian Li, Ruqi Zhang<\/a> and Yingzhen Li<\/strong>. Bayesian Computation in Deep Learning<\/a>. Handbook of Markov Chain Monte Carlo, 2nd Edition (in press, expected 2025\/26, refereed chapter)<\/p>\n

Zijing Ou<\/a>, Mingtian Zhang, Andi Zhang, Tim Z. Xiao, Yingzhen Li<\/strong> and David Barber<\/a>. Improving Probabilistic Diffusion Models With Optimal Diagonal Covariance Matching<\/a>. International Conference on Learning Representations (ICLR), 2025. (oral, 1.5%)<\/p>\n

Tobias Schr\u00f6<\/span>der, Zijing Ou<\/a>, Yingzhen Li<\/strong> and Andrew Duncan<\/a>. Energy-Based Modelling for Discrete and Mixed Data via Heat Equations on Structured Spaces<\/a>. Neural Information Processing Systems (NeurIPS), 2024.<\/p>\n

Carles Balsells-Rodas<\/a>, Yixin Wang<\/a> and Yingzhen Li<\/strong>. On the Identifiability of Switching Dynamical Systems<\/a>. International Conference on Machine Learning (ICML), 2024.<\/p>\n

Papamakou et al. Position: Bayesian Deep Learning is Needed in the Age of Large-Scale AI<\/a>. International Conference on Machine Learning (ICML), 2024.<\/p>\n

Hee Suk Yoon, Eunseop Yoon, Joshua Tian Jin Tee, Mark A. Hasegawa-Johnson, Yingzhen Li<\/strong> and Chang D. Yoo<\/a>. C-TPT: Calibrated Test-Time Prompt Tuning for Vision-Language Models via Text Feature Dispersion<\/a>. International Conference on Learning Representations (ICLR), 2024.<\/p>\n

Tobias Schr\u00f6<\/span>der, Zijing Ou<\/a>, Jen Ning Lim, Yingzhen Li<\/strong>, Sebastian Vollmer and Andrew Duncan<\/a>.\u00a0<\/span>Energy Discrepancies: A Score-Independent Loss for Energy-Based Models<\/a>. Neural Information Processing Systems (NeurIPS), 2023.<\/p>\n

Harrison Zhu<\/a>, Carles Balsells Rodas<\/a> and Yingzhen Li<\/strong>. Markovian Gaussian Process Variational Autoencoders<\/a>. International Conference on Machine Learning (ICML), 2023.<\/p>\n

Wenlong Chen<\/a> and Yingzhen Li<\/strong>. Calibrating Transformers via Sparse Gaussian Processes<\/a>. International Conference on Learning Representations (ICLR),\u00a02023.<\/p>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
\n

Hee Suk Yoon<\/a>, Joshua Tian Jin Tee, Gwangsu Kim, Eunseop Yoon, Sunjae Yoon, Yingzhen Li<\/strong> and Chang D. Yoo<\/a>. ESD: Expected Squared Difference as a Tuning-Free Trainable Calibration Measure<\/a>. International Conference on Learning Representations (ICLR),\u00a02023.<\/p>\n<\/div>\n<\/div>\n<\/div>\n

Ryutaro Tanno<\/a>, Melanie F. Pradier<\/a>, Aditya Nori<\/a> and Yingzhen Li<\/strong>. Repairing Neural Networks by Leaving the Right Past Behind<\/a>. Neural Processing Information Systems (NeurIPS),\u00a02022.<\/p>\n

Yanzhi Chen<\/a>, Weihao Sun, Yingzhen Li<\/strong> and Adrian Weller<\/a>. Scalable Infomin Learning<\/a>. Neural Processing Information Systems (NeurIPS),\u00a02022.<\/p>\n

Zijing Ou<\/a>, Tingyang Xu<\/a>, Qinliang Su<\/a>, Yingzhen Li<\/strong>, Peilin Zhao<\/a> and Yatao Bian<\/a>. Learning Set Functions Under the Optimal Subset Oracle via Equivariant Variational Inference<\/a>. Neural Processing Information Systems (NeurIPS),\u00a02022.<\/p>\n

Hippolyt Ritter<\/a>, Martin Kukla<\/a>, Cheng Zhang<\/a> and Yingzhen Li<\/strong>. Sparse Uncertainty Representation in Deep Learning with Inducing Weights<\/a>. Neural Processing Information Systems (NeurIPS), 2021.<\/p>\n

Thomas Henn<\/a>, Yasukazu Sakamoto<\/a>, Cl\u00e9ment Jacquet<\/a>, Shunsuke Yoshizawa<\/a>, Masamichi Andou<\/a>, Stephen Tchen<\/a>, Ryosuke Saga<\/a>, Hiroyuki Ishihara<\/a>, Katsuhiko Shimizu<\/a>, Yingzhen Li<\/strong>\u00a0and\u00a0Ryutaro Tanno<\/a>.\u00a0A Principled Approach to Failure Analysis and Model Repairment: Demonstration in Medical Imaging<\/a>. International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2021.<\/p>\n

Wenbo Gong<\/a>, Kaibo Zhang, Yingzhen Li<\/strong> and Jos\u00e9 Miguel Hern\u00e1ndez-Lobato<\/a>. Active Slices for Slided Stein Discrepancy<\/a>. International Conference on Machine Learning (ICML),\u00a02021<\/p>\n

Wenbo Gong<\/a>, Yingzhen Li<\/strong> and Jos\u00e9 Miguel Hern\u00e1ndez-Lobato<\/a>. Sliced Kernelized Stein Discrepancy<\/a>. International Conference on Learning Representations (ICLR), 2021.<\/p>\n

Ruqi Zhang<\/a>, Yingzhen Li<\/strong>, Chris De Sa<\/a>, Sam Devlin<\/a> and Cheng Zhang<\/a>.\u00a0Meta-Learning for Variational Inference<\/a>. International Conference on Artificial Intelligence and Statistics (AISTATS), 2021.<\/p>\n

\n
\n
\n

Yi Zhu<\/a>, Ehsan Shareghi<\/a>, Yingzhen Li<\/strong>, Roi Reichart<\/a> and Anna Korhonen<\/a>. Combining Deep Generative Models and Multi-lingual Pretraining for Semi-supervised Document Classification<\/a>.\u00a0European Chapter of the Association for Computational Linguistics (EACL), 2021.<\/p>\n<\/div>\n<\/div>\n<\/div>\n

Andrew Y. K. Foong<\/a>*,\u00a0David R. Burt<\/a>*,\u00a0Yingzhen Li<\/strong>\u00a0and\u00a0Richard E. Turner<\/a>.\u00a0On the Expressiveness of Approximate Inference in Bayesian Neural Networks<\/a>. Neural Processing Information Systems (NeurIPS), 2020.<\/p>\n

Cheng Zhang<\/a>, Kun Zhang<\/a> and Yingzhen Li<\/strong>. A Causal View on Robustness of Neural Networks<\/a>. Neural Processing Information Systems (NeurIPS), 2020.<\/p>\n

Maximilian Igl<\/a>,\u00a0Kamil Ciosek<\/a>,\u00a0Yingzhen Li<\/strong>,\u00a0Sebastian Tschiatschek<\/a>,\u00a0Cheng Zhang<\/a>,\u00a0Sam Devlin<\/a> and\u00a0Katja Hofmann<\/a>.\u00a0Generalization in Reinforcement Learning with Selective Noise Injection and Information Bottleneck<\/a>. Neural Processing Information Systems (NeurIPS), 2019.<\/p>\n

Ehsan Shareghi<\/a>, Yingzhen Li<\/strong>, Yi Zhu<\/a>, Roi Reichart<\/a> and Anna Korhonen<\/a>.\u00a0Bayesian Learning for Neural Dependency Parsing<\/a>. NAACL-HLT 2019.<\/p>\n

Chao Ma<\/a>, Yingzhen Li<\/strong> and Jos\u00e9 Miguel Hern\u00e1ndez-Lobato<\/a>. Variational Implicit Processes<\/a>. International Conference on Machine Learning (ICML), 2019. code<\/a><\/span><\/p>\n

Yingzhen Li<\/strong>, John Bradshaw<\/a> and Yash Sharma<\/a>.\u00a0Are Generative Classifiers More Robust to Adversarial Attacks?<\/a>\u00a0International Conference on Machine Learning (ICML), 2019. code<\/a>\u00a0cartoon<\/a><\/span>\n

<\/span><\/span><\/p>\n

\"\"<\/a><\/p>\n

(Generative\/Discriminative classifiers, the graphical model defines its nature)<\/span><\/p>\n

\"\"<\/a><\/p>\n

(Fusing generative\/discriminative behaviours: combining VGG-features and generative models)<\/span><\/p>\n

Adversarial machine learning research has become an arm race, so far the adversary side is taking the lead. We conjecture in this paper that feed-forward DNN's vulnerability to adversarial examples is due to its discriminative nature, so instead we build deep generative models and turn them into classifiers using Bayes' rule<\/strong>. Extensive empirical study shows advantages of generative classifiers over discriminative ones, both in terms of robustness and detection to adversarial examples. We also build a generative-discriminative hybrid for natural image classification, which combines the best from both worlds.<\/span><\/p>\n

<\/div><\/p>\n

Wenbo Gong<\/a>*, Yingzhen Li<\/strong>* and Jos\u00e9 Miguel Hern\u00e1ndez-Lobato<\/a>. Meta-Learning for Stochastic Gradient MCMC<\/a>. International Conference on Learning Representations (ICLR), 2019. \u00a0code<\/a><\/span>\u00a0cartoon<\/a><\/span>\n

<\/p>\n

\"\"<\/a><\/p>\n

(finding the best one from the largest set of provably correct samplers\u00a0?)<\/span><\/p>\n

\"\"<\/a><\/p>\n

(the meta-learned sampler is trained on an isotropic Gaussian distribution)<\/span><\/p>\n

Recently people are very excited about meta-learning<\/a> algorithms<\/a> that are then used to optimise ML models. However naively proposing an arbitrary neural network to do sampling<\/em> is almost guaranteed to fail. So we really want to meta-learn an MCMC sampler that is provably correct (i.e. the stationary distribution is the target distribution), and we build on top of a seminal work on SG-MCMC<\/a> to combine the flexibility of neural networks and Hamiltonian Monte Carlo.<\/span><\/p>\n

<\/div><\/p>\n

Yingzhen Li<\/strong> and Stephan Mandt<\/a>.\u00a0Disentangled Sequential Autoencoder<\/a>. International Conference on Machine Learning (ICML),\u00a02018. sprites data<\/span><\/a>\u00a0architecture<\/span><\/a>\u00a0cartoon<\/a><\/span>\n

<\/p>\n

\"\"<\/a><\/p>\n

(embarrassingly simple\u00a0?<\/span>)<\/span><\/p>\n

While there are quite a few recent papers talking about learning disentangled representation using some new loss function, I thought: why not just go back to the graphical model land, and directly enforce disentanglement in model design? So there you go a minimalistic generative model for disentangling content and dynamics in a sequence. We have speech recognition example in the paper as well, and we also show using stochastic dynamics could potentially achieve better compression.<\/span><\/p>\n

Part of this project was done as an intern at Disney research.<\/span><\/p>\n

(Previously titled \"A Deep Generative Model for Disentangled Representations of Sequential Data\")<\/span><\/p>\n

<\/div><\/p>\n

Cuong V. Nguyen<\/a>, Yingzhen Li<\/strong>, Thang D. Bui<\/a> and Richard E. Turner<\/a>.\u00a0Variational Continual Learning<\/a>. International Conference on Learning Representations (ICLR), 2018. \u00a0code<\/a> <\/span> cartoon<\/a><\/span>\n

<\/p>\n

\"\"<\/a>(Well, quite a bit exaggeration actually, we need to make it practical ?)<\/span><\/p>\n

\"Bayesian Inference is **the natural choice** for online\/incremental\/continual learning.\" Conceptually this is nice, but you know deep learning engineers wouldn't bid for it as exact Bayesian inference takes \"forever\" to run. In this paper we just use variational Bayes to make it tractable and easy to implement for continual learning of deep learning tasks. For both supervised and unsupervised learning.<\/span><\/p>\n

<\/div><\/p>\n

Yingzhen Li<\/strong> and Richard E. Turner<\/a>. Gradient Estimators for Implicit Models<\/a>. International Conference on Learning Representations (ICLR), 2018. \u00a0code<\/a> <\/span> cartoon<\/a><\/span>\n

<\/p>\n

\"\"<\/a><\/p>\n

(So why not try gradient approximations instead ?)<\/span><\/p>\n

One of the main reasons why people now are super excited about<\/span> GANs<\/a> is that the generative model is implicitly specified by a neural network warping simple noise vectors. However, applying probabilistic learning methods to these models requires severe approximations, where an overfitting estimation to the optimisation objective can lead to arbitrarily bad models. In this work, we explore an alternative route -- approximating the gradient function -- for training implicit models. We demonstrate the efficacy of our method with some exciting applications: meta-learning for posterior samplers, and improving GAN's sample diversity.<\/span><\/p>\n

<\/div><\/p>\n

Yingzhen Li<\/strong> and Yarin Gal<\/a>. Dropout Inference in Bayesian Neural Networks with Alpha-divergences<\/a>. International Conference on Machine Learning (ICML), 2017. \u00a0code<\/a><\/span> cartoon<\/a><\/span>\n

<\/p>\n

\"bbalpha_dropout\"<\/a> (Code for the core idea in<\/span> Keras<\/a>. With other helper functions dropout+BB-alpha can be implemented in less than 30 lines.)<\/span><\/p>\n

\"\"<\/a><\/p>\n

(Bayesian NN could detect adversarial examples (FGS targeted attack))<\/span><\/p>\n

Show me your loss function, and I will return you an objective function that trains a Bayesian NN with<\/span> BB-alpha<\/a> and dropout! We extended the<\/span> dropout variational inference<\/a> idea to minimising alpha-divergences, and the proposed re-formulation can be implemented very efficiently with any high-level deep learning framework. You can try any dropout techniques (<\/span>Bernoulli, Gaussian<\/a>, Concrete<\/a>, dropConnect<\/a>, and even<\/span> stochastic depth<\/a>) and select the best alpha for your need. As a side-project we also showed that Bayesian NNs can be used to detect adversarial examples, both in untargeted and targeted attacks.<\/span><\/p>\n

<\/div><\/p>\n

Yingzhen Li<\/strong> and Richard E. Turner<\/a>. R\u00e9nyi Divergence Variational Inference<\/a>. Neural Processing Information Systems (NIPS), 2016.\u00a0code<\/a><\/span>\u00a0 cartoon<\/a><\/span>\n

<\/p>\n

\"emoji_vrbound\"<\/a><\/p>\n

(credit to<\/span> Ferenc Husz\u00e1r<\/a>, thanks for your fun emoji illustration!)<\/span><\/p>\n

\"meanfield_vr\"<\/a><\/p>\n

(mean-field approximation for Bayesian linear regression)<\/span><\/p>\n

The goal of approximate inference is to get accurate posterior approximations in a fast way. However the definition of \"accurate\" really depends on what you actually want. You as the user specify the approximation family that suits your constraints, and my job here is to provide you a flexible inference framework that allows you to fine-tune the alpha parameter to get the best for your need.<\/strong> No painstaking derivations again and again!<\/span><\/p>\n

(Previously titled \"Variational Inference with R\u00e9nyi Divergence\")<\/span><\/p>\n

<\/div><\/p>\n

Jos\u00e9 Miguel Hern\u00e1ndez-Lobato<\/a>*, Yingzhen Li<\/strong>*, Mark Rowland<\/a>, Daniel Hern\u00e1ndez-Lobato<\/a>, Thang Bui<\/a>\u00a0and\u00a0Richard E. Turner<\/a>. Black-box \u03b1<\/span><\/span><\/span><\/span><\/span><\/span>-divergence Minimization<\/a>. International Conference on Machine Learning (ICML), 2016.\u00a0 code<\/a><\/span> cartoon<\/a><\/span>\n

<\/p>\n

\"alpha_divergence\"<\/a><\/p>\n

A taste of alpha-divergences. (Original visualisation credit:<\/span> Tom Minka<\/a>)<\/span><\/p>\n

\"relationship\"<\/a><\/p>\n

Connections to existing methods.<\/span><\/p>\n

<\/div><\/p>\n

Thang Bui<\/a>, Daniel Hern\u00e1ndez-Lobato<\/a>, Jos\u00e9 Miguel Hern\u00e1ndez-Lobato<\/a>, Yingzhen Li<\/strong> and\u00a0Richard E. Turner<\/a>. Deep Gaussian Processes for Regression using Approximate Expectation Propagation<\/a>. International Conference on Machine Learning (ICML), 2016.\u00a0 code<\/a> <\/span>cartoon<\/a><\/span>\n

<\/p>\n

\"deepgp_cartoon\"<\/a><\/p>\n

(Fig credit: Wikipedia pages & Thang Bui<\/a>)<\/p>\n

Under<\/span> some conditions<\/a>, a Gaussian process model can be viewed as a one-hidden-layer neural network with an infinite number of hidden units<\/strong>. So it makes sense to make GPs \"deep\", and there you go the<\/span> deep Gaussian processes<\/a>. However inference for deep GPs is non-trivial unlike regular deep neural networks, so this paper introduces a scalable inference method<\/strong> for them using Gaussian propagation. GPs are state-of-the-art regression models and our contribution make them even so!<\/span><\/p>\n

<\/div><\/p>\n

Yingzhen Li<\/strong>, Jose Miguel Hernandez-Lobato<\/a> and Richard E. Turner<\/a>. Stochastic Expectation Propagation<\/a>.<\/em>\u00a0Neural Processing Information Systems (NIPS), 2015 (spotlight, 4.5%<\/strong>).\u00a0 demo<\/a><\/span> cartoon<\/a><\/span>\n

<\/p>\n

\"sep\"<\/a><\/p>\n

<\/div><\/p>\n

Workshop Preprints<\/strong><\/span><\/p>\n

Sebastian Lunz<\/a>, Yingzhen Li<\/strong>, Andrew Fitzgibbon<\/a> and Nate Kushman<\/a>. Inverse Graphics GAN: Learning to Generate 3D Shapes from Unstructured 2D Data<\/a>. NeurIPS 2020 workshop on\u00a0Differentiable vision, graphics, and physics applied to machine learning (DiffCVGP), 2020.<\/p>\n

Chaochao Lu<\/a>, Richard E. Turner<\/a>, Yingzhen Li<\/strong> and Nate Kushman<\/a>. Interpreting Spatially Infinite Generative Models<\/a>. ICML 2020 Workshop on Human Interpretability in Machine Learning (WHI), 2020.<\/p>\n

Andrew Y.K. Foong<\/a>, Yingzhen Li<\/strong>,\u00a0Jos\u00e9 Miguel Hern\u00e1ndez-Lobato<\/a> and Richard E. Turner<\/a>. \"In-Between\" Uncertainty for Bayesian Neural Networks<\/a>. ICML 2019 workshop on Uncertainty & Robustness in Deep Learning (oral)<\/p>\n

Yingzhen Li<\/strong>. Approximate Gradient Descent for Training Implicit Generative Models<\/a>. NIPS 2017 Bayesian Deep Learning workshop. 2017.<\/p>\n

Yingzhen Li<\/strong>, Richard E. Turner<\/a> and Qiang Liu<\/a>. Approximate Inference with Amortised MCMC<\/a>. ICML 2017 Workshop on Implicit Generative Models. 2017. cartoon<\/a><\/span>\n

<\/p>\n

\"amccartoon\"<\/a><\/p>\n

(Yes, this is the whole idea, no much need to explain in equations.)<\/span><\/p>\n

Amortised MCMC is an extremely flexible approximate inference framework. It is completely up to you to specify the sample generator, the Markov chain transition kernel, and the update rule for that generator. In the experiments, we specifically defined the sample generating process by warping simple randomness through a deep neural network, and the update rule was a<\/span> GAN<\/a>-like method but acting in latent space. I'm very excited about this work and would like to see extensions to discrete distributions as well!<\/span><\/p>\n

<\/div><\/p>\n

Yingzhen Li<\/strong> and Qiang Liu<\/a>. Wild Variational Approximations<\/a>. preprint presented in NIPS Advances in approximate inference, 2016. cartoon<\/a><\/span>\n

<\/p>\n

\"zoo\"<\/a>(A zoo of inference engines available at late 2016)<\/span><\/p>\n

It would be fantastic if we can get the \"lion king\" (neural network transform) to help beat the \"demon dragon\" (exact posterior)! In this preliminary work we discussed several potential proposals to train a \"wild approximation\" in which prediction is tractable but the density of q might be difficult to evaluate. We presented it at the NIPS 2016 approximate inference workshop with preliminary results for toy examples and generative models.<\/span><\/p>\n

<\/div><\/p>\n

Yingzhen Li<\/strong> and Richard E. Turner<\/a>. A Unifying Approximate Inference Framework from Variational Free Energy Relaxation<\/a>. NIPS Advances in approximate inference workshop, 2016<\/p>\n

Daniel Hern\u00e1ndez-Lobato<\/a>, Jos\u00e9 Miguel Hern\u00e1ndez-Lobato<\/a>, Yingzhen Li<\/strong>, Thang Bui<\/a>\u00a0and\u00a0Richard E. Turner<\/a>. Stochastic Expectation Propagation for Large Scale Gaussian Process Classification<\/a>. NIPS Advances in approximate inference workshop (contributed talk), 2015<\/p>\n

Yingzhen Li<\/strong> and Ye Zhang. Generating ordered list of Recommended Items: a Hybrid Recommender System of Microblog<\/a>. 2012<\/p>\n

Thesis<\/strong><\/p>\n

Approximate Inference: New Visions.<\/a> PhD in Engineering, University of Cambridge, June 2018.<\/p>\n

Compressed Sensing and Related Learning Problems.<\/a><\/em> B.S. in Mathematics, Sun Yat-sen University, May 2013.<\/em> (Best B.S. Thesis Award).\u00a0[slides<\/a>]<\/p>\n

Things I've done in my undergrad years<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

Working papers Wuhao Chen, Zijing Ou and Yingzhen Li. Neural Flow Samplers with Shortcut Models. In submission. 2025. Harrison Zhu, Adam Howes, Owen van Eer, Maxime Richard, Yingzhen Li, Dino Sejdinovic and Seth Flaxman. Aggregated Gaussian Processes with Multiresolution Earth Observation Covariates. 2022. Wenbo Gong and Yingzhen Li.\u00a0Interpreting Diffusion Score Matching using Normalizing Flow. ICML ... [Read more...]<\/a><\/span><\/p>\n","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"open","ping_status":"open","template":"","meta":{"footnotes":""},"class_list":["entry","page","publish","author-admin","post-345"],"_links":{"self":[{"href":"https:\/\/yingzhenli.net\/home\/en\/index.php?rest_route=\/wp\/v2\/pages\/345","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/yingzhenli.net\/home\/en\/index.php?rest_route=\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/yingzhenli.net\/home\/en\/index.php?rest_route=\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/yingzhenli.net\/home\/en\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/yingzhenli.net\/home\/en\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=345"}],"version-history":[{"count":230,"href":"https:\/\/yingzhenli.net\/home\/en\/index.php?rest_route=\/wp\/v2\/pages\/345\/revisions"}],"predecessor-version":[{"id":1719,"href":"https:\/\/yingzhenli.net\/home\/en\/index.php?rest_route=\/wp\/v2\/pages\/345\/revisions\/1719"}],"wp:attachment":[{"href":"https:\/\/yingzhenli.net\/home\/en\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=345"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}