Neil Houlsby
Neil is a Senior Research Scientist in the Google Brain team, Zürich. He works on Machine Learning, in particular on transfer learning, representation learning, AutoML, computer vision, and NLP. Previous he received his PhD from the Cambridge CBL Laboratory.
Research Areas
Authored Publications
Google Publications
Other Publications
Sort By
PaLI-X: On Scaling up a Multilingual Vision and Language Model
Josip Djolonga
Piotr Padlewski
Basil Mustafa
Carlos Riquelme
Yi Tay
Siamak Shakeri
Daniel Salz
Michael Tschannen
Mandar Joshi
Filip Pavetić
Anurag Arnab
Yuanzhong Xu
Keran Rong
Computer Vision and Pattern Recognition Conference (CVPR) (2024)
Preview abstract
We explore the boundaries of scaling up a multilingual vision and language model, both in terms of size of the components and the breadth of its training task mixture. Our model achieves new levels of performance on a wide-range of varied and complex tasks, including multiple image-based captioning and question-answering tasks, image-based document understanding and few-shot (in-context) learning, as well as object detection, video question answering, and video captioning. Our model advances the state-of-the-art on most vision-and-language benchmarks considered (20+ of them). Finally, we observe emerging capabilities, such as complex counting and multilingual object detection, tasks that are not explicitly in the training mix.
View details
UL2: Unifying Language Learning Paradigms
Yi Tay
Xavier Garcia
Jason Wei
Hyung Won Chung
Steven Zheng
ICLR (2023)
Preview abstract
Existing pre-trained models are generally geared towards a particular class of
problems. To date, there seems to be still no consensus on what the right architecture and pre-training setup should be. This paper presents a unified framework for
pre-training models that are universally effective across datasets and setups. We
begin by disentangling architectural archetypes with pre-training objectives – two
concepts that are commonly conflated. Next, we present a generalized and unified perspective for self-supervision in NLP and show how different pre-training
objectives can be cast as one another and how interpolating between different
objectives can be effective. We then propose Mixture-of-Denoisers (MoD), a pretraining objective that combines diverse pre-training paradigms together. We furthermore introduce a notion of mode switching, wherein downstream fine-tuning
is associated with specific pre-training schemes. We conduct extensive ablative
experiments to compare multiple pre-training objectives and find that our method
pushes the Pareto-frontier by outperforming T5 and/or GPT-like models across
multiple diverse setups. Finally, by scaling our model up to 20B parameters, we
achieve SOTA performance on 50 well-established supervised NLP tasks ranging from language generation (with automated and human evaluation), language
understanding, text classification, question answering, commonsense reasoning,
long text reasoning, structured knowledge grounding and information retrieval.
Our model also achieve strong results at in-context learning, outperforming 175B
GPT-3 on zero-shot SuperGLUE and tripling the performance of T5-XXL on oneshot summarization. Finally, we show that UL2 20B works well with chain-ofthought prompting and reasoning tasks, making it an appealing choice for research
into reasoning at a small to medium scale of 20B parameters. We publicly release
Flax-based T5X model checkpoints for the 20B model.
View details
Scaling Vision Transformers to 22 Billion Parameters
Josip Djolonga
Basil Mustafa
Piotr Padlewski
Justin Gilmer
Mathilde Caron
Rodolphe Jenatton
Michael Tschannen
Anurag Arnab
Carlos Riquelme
Fisher Yu
Avital Oliver
Fantine Huot
Mark Collier
Yi Tay
Filip Pavetić
Thomas Kipf
Arxiv (2023)
Preview abstract
The scaling of Transformers has driven breakthrough capabilities for language models.
At present, the largest large language models (LLMs) contain upwards of 100B parameters.
Vision Transformers (ViT) have introduced the same architecture to image and video modeling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters. We present a recipe for highly efficient training of a 22B-parameter ViT and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features) ViT22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between bias and performance, an improved alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT22B demonstrates the potential for "LLM-like'' scaling in vision, and provides key steps towards getting there.
View details
Dual PatchNorm
Transactions on Machine Learning Research (2023) (to appear)
Preview abstract
We discover that just placing two LayerNorms: before and after the patch embedding layer leads to improvements over well-tuned ViT models. In particular, this outperforms exhaustive search for alternative LayerNorm placement strategies in the transformer block itself.
View details
PaLI: A Jointly-Scaled Multilingual Language-Image Model
Piotr Padlewski
Daniel Salz
Basil Mustafa
Keran Rong
Hassan Akbari
Linting Xue
James Bradbury
Carlos Riquelme
International Conference on Learning Representations (ICLR) (2023)
Preview abstract
Effective scaling and a flexible task interface enable large-capacity language models to excel at many tasks. PaLI (Pathways Language and Image model) extends these ideas to the joint modeling of language and vision. PaLI is a model that generates text based on visual and textual inputs. Using this API, PaLI is able to perform many vision, language, and multimodal tasks, across many languages. We train PaLI with two main principles: reuse of pretrained unimodal components, and joint scaling of modalities. Using large-capacity pretrained language models and vision models allows us to capitalize on their existing capabilities, while leveraging the substantial cost of training them. We scale PaLI models across three axes:the language component, the vision component, and the training data that fuses them. For the vision component, we train the largest and best-performing VisionTransformer (ViT) to date. For the data, we build an image-text training set over10B images and covering over 100 languages.
PaLI inherits and enhances language-understanding capabilities, and achieves state-of-the-art in multiple vision and language tasks (image classification, image captioning, visual question-answering, scene-text understanding, etc.), based on a simple, modular, and reuse-friendly platform for modeling and scaling.
View details
Robust and data-efficient generalization of self-supervised machine learning for diagnostic imaging
Laura Anne Culp
Jan Freyberg
Basil Mustafa
Sebastien Baur
Simon Kornblith
Ting Chen
Patricia MacWilliams
Sara Mahdavi
Megan Zoë Walker
Aaron Loh
Cameron Chen
Scott Mayer McKinney
Zach William Beaver
Fiona Keleher Ryan
Mozziyar Etemadi
Umesh Telang
Lily Hao Yi Peng
Geoffrey Everest Hinton
Mohammad Norouzi
Nature Biomedical Engineering (2023)
Preview abstract
Machine-learning models for medical tasks can match or surpass the performance of clinical experts. However, in settings differing from those of the training dataset, the performance of a model can deteriorate substantially. Here we report a representation-learning strategy for machine-learning models applied to medical-imaging tasks that mitigates such ‘out of distribution’ performance problem and that improves model robustness and training efficiency. The strategy, which we named REMEDIS (for ‘Robust and Efficient Medical Imaging with Self-supervision’), combines large-scale supervised transfer learning on natural images and intermediate contrastive self-supervised learning on medical images and requires minimal task-specific customization. We show the utility of REMEDIS in a range of diagnostic-imaging tasks covering six imaging domains and 15 test datasets, and by simulating three realistic out-of-distribution scenarios. REMEDIS improved in-distribution diagnostic accuracies up to 11.5% with respect to strong supervised baseline models, and in out-of-distribution settings required only 1–33% of the data for retraining to match the performance of supervised models retrained using all available data. REMEDIS may accelerate the development lifecycle of machine-learning models for medical imaging.
View details
Preview abstract
Human-like perceptual similarity is an emergent property in the intermediate feature space of ImageNet-pretrained classifiers. Perceptual distances between images, as measured in the space of pre-trained image embeddings, have outperformed prior low-level metrics significantly on assessing image similarity. This has led to the wide adoption of perceptual distances as both an evaluation metric and an auxiliary training objective for image synthesis tasks. While image classification has improved by leaps and bounds, the de facto standard for computing perceptual distances uses older, less accurate models such as VGG and AlexNet. Motivated by this, we evaluate the perceptual scores of modern networks: ResNets, EfficientNets and VisionTransformers. Surprisingly, we observe an inverse correlation between ImageNet accuracy and perceptual scores: better classifiers achieve worse perceptual scores. We dive deeper into this, studying the ImageNet accuracy/perceptual score relationship under different hyperparameter configurations. Improving accuracy improves perceptual scores up to a certain point, but beyond this point we uncover a Pareto frontier between accuracies and perceptual scores. We explore this relationship further using distortion invariance, spatial frequency sensitivity and alternative perceptual functions. Based on our study, we find a ImageNet trained ResNet-6 network whose emergent perceptual score matches the best prior score obtained with networks trained explicitly on a perceptual similarity task.
View details
Simple Open-Vocabulary Object Detection with Vision Transformers
Austin Stone
Maxim Neumann
Dirk Weissenborn
Alexey Dosovitskiy
Anurag Arnab
Zhuoran Shen
Thomas Kipf
ECCV (Poster) (2022)
Preview abstract
Combining simple architectures with large-scale pre-training has led to massive improvements in image classification. For object detection, pre-training and scaling approaches are less well established, especially in the long-tailed and open-vocabulary setting, where training data is relatively scarce. In this paper, we propose a strong recipe for transferring image-text models to open-vocabulary object detection. We use a standard Vision Transformer architecture with minimal modifications, contrastive image-text pre-training, and end-to-end detection fine-tuning. Our analysis of the scaling properties of this setup shows that increasing image-level pre-training and model size yield consistent improvements on the downstream detection task. We provide the adaptation strategies and regularizations needed to attain very strong performance on zero-shot text-conditioned and one-shot image-conditioned object detection. Code and models are available on GitHub (https://github.com/google-research/scenic/tree/main/scenic/projects/owl_vit).
View details
Preview abstract
Attention-based neural networks such as the Vision Transformer (ViT) have recently attained state-of-the-art results on many computer vision benchmarks. Scale is a primary ingredient in attaining excellent results, therefore, understanding a model's scaling properties is a key to designing future generations effectively. While the laws for scaling Transformer language models have been studied, it is unknown how Vision Transformers scale. To address this, we scale ViT models and data, both up and down, and characterize the relationships between error rate, data, and compute. Along the way, we refine the architecture and training of ViT, reducing memory consumption and increasing accuracy of the resulting models. As a result, we successfully train a ViT model with two billion parameters, which attains a new state-of-the-art on ImageNet of 90.45% top-1 accuracy. The model also performs well for few-shot transfer, for example, reaching 84.86% top-1 accuracy on ImageNet with only 10 examples per class.
View details
On Robustness and Transferability of Convolutional Neural Networks
Josip Djolonga
Jessica Yung
Michael Tschannen
Rob Romijnders
Dan Moldovan
Sylvain Gelly
Conference on Computer Vision and Pattern Recognition (2021)
Preview abstract
Modern deep convolutional networks (CNNs) are often criticized for their failure to generalize under distributional shifts. However, several recent breakthroughs in transfer learning suggest that these networks can cope with severe distribution shifts and successfully adapt to new tasks from a few training examples. In this work we revisit the out-of-distribution and transfer performance of modern image classification CNNs and investigate the impact of the pre-training data scale, the model scale, and the data preprocessing pipeline. We find that increasing both the training set and model sizes significantly improve the robustness to distribution shifts. Furthermore, we show that, perhaps surprisingly, simple changes in the preprocessing such as modifying the image resolution can significantly mitigate robustness issues in some cases. Finally, we outline the shortcomings of existing robustness evaluation datasets and introduce a synthetic dataset for fine-grained robustness analysis.
View details
Supervised Transfer Learning at Scale for Medical Imaging
Aaron Loh
Basil Mustafa
Jan Freyberg
Patricia MacWilliams
Megan Wilson
Scott Mayer McKinney
Peggy Bui
Umesh Telang
ArXiV (2021)
Preview abstract
Transfer learning is a standard building block of successful medical imaging models, yet previous efforts suggest that at limited scale of pre-training data and model capacity, benefits of transfer learning to medical imaging are insubstantial. In this work, we explore whether scaling up pre-training can help improve transfer to medical tasks. In particular, we show that when using the Big Transfer recipe to further scale up pre-training, we can indeed considerably improve transfer performance across three popular yet diverse medical imaging tasks - interpretation of chest radiographs, breast cancer detection from mammograms and skin condition detection from smartphone images. Despite pre-training on unrelated source domains, we show that scaling up the model capacity and pre-training data yields performance improvements regardless of how much downstream medical data is available. In particular, we show suprisingly large improvements to zero-shot generalisation under distribution shift. Probing and quantifying other aspects of model performance relevant to medical imaging and healthcare, we demonstrate that these gains do not come at the expense of model calibration or fairness.
View details
A Unified Few-Shot Classification Benchmark to Compare Transfer and Meta Learning Approaches
Sylvain Gelly
NeurIPS Datasets and Benchmarks Track (2021)
Preview abstract
Meta and transfer learning are two successful families of approaches to few-shot learning. Despite highly related goals, state-of-the-art advances in each family are measured largely in isolation of each other. As a result of diverging evaluation norms, a direct or thorough comparison of different approaches is challenging. To bridge this gap, we introduce a few-shot classification evaluation protocol named VTAB+MD with the explicit goal of facilitating sharing of insights from each community. We demonstrate its accessibility in practice by performing a cross-family study of the best transfer and meta learners which report on both a large-scale meta-learning benchmark (Meta-Dataset, MD), and a transfer learning benchmark (Visual Task Adaptation Benchmark, VTAB). We find that, on average, large-scale transfer methods (Big Transfer, BiT) outperform competing approaches on MD, even when trained only on ImageNet. In contrast, meta-learning approaches struggle to compete on VTAB when trained and validated on MD. However, BiT is not without limitations, and pushing for scale does not improve performance on highly out-of-distribution MD tasks. We hope that this work contributes to accelerating progress on few-shot learning research.
View details
Revisiting the Calibration of Modern Neural Networks
Josip Djolonga
Rob Romijnders
Frances Ann Hubis
Neural Information Processing Systems (2021) (to appear)
Preview abstract
Accurate estimation of predictive uncertainty (model calibration) is essential for the safe application of neural networks. Many instances of miscalibration in modern neural networks have been reported, suggesting a trend that newer, more accurate models produce poorly calibrated predictions. Here, we revisit this question for recent state-of-the-art image classification models. We systematically relate model calibration and accuracy, and find that the most recent models, notably those not using convolutions, are among the best calibrated. Trends observed in prior model generations, such as decay of calibration with distribution shift or model size, are less pronounced in recent architectures. We also show that model size and amount of pretraining do not fully explain these differences, suggesting that architecture is a major determinant of calibration properties.
View details
MLP-Mixer: An All-MLP Architecture for Vision
Jessica Yung
Jakob Uszkoreit
Alexey Dosovitskiy
NeurIPS 2021 (poster)
Preview abstract
Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e. "mixing" the per-location features), and one with MLPs applied across patches (i.e. "mixing" spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks with comparable pre-training and inference cost. We hope that these results spark further research beyond the realms of well established CNNs and Transformers.
View details
Scalable Transfer Learning with Expert Models
Carlos Riquelme
Basil Mustafa
Cedric Renggli
André Susano Pinto
Sylvain Gelly
ICLR 2021 (2021)
Preview abstract
Transfer of pre-trained representations can improve sample efficiency and reduce computational requirements for new tasks. However, representations used for transfer are usually generic, and are not tailored to a particular distribution of downstream tasks. We explore the use of expert representations for transfer with a simple, yet effective, strategy. We train a diverse set of experts by exploiting existing label structures, and use cheap-to-compute performance proxies to select the relevant expert for each target task. This strategy scales the process of transferring to new tasks, since it does not revisit the pre-training data during transfer. Accordingly, it requires little extra
compute per target task, and results in a speed-up of 2–3 orders of magnitude compared to competing approaches. Further, we provide an adapter-based architecture able to compress many experts into a single model. We evaluate our approach on two different data sources and demonstrate that it outperforms baselines on over 20 diverse vision tasks in both cases.
View details
Representation learning from videos in-the-wild: An object-centric approach
Rob Romijnders
Michael Tschannen
Josip Djolonga
WACV (2021)
Preview abstract
We propose a method to learn image representations from uncurated videos. We combine a supervised loss from off-the-shelf object detectors and self-supervised losses which naturally arise from the video-shot-frame-object hierarchy present in each video. We report competitive results on 19 transfer learning tasks of the Visual Task Adaptation Benchmark (VTAB), and on 8 out-of-distribution-generalization tasks, and discuss the benefits and shortcomings of the proposed approach. In particular, it improves over the baseline on all 18/19 few-shot learning tasks and 8/8 out-of-distribution generalization tasks. Finally, we perform several ablation studies and analyze the impact of the pretrained object detector on the performance across this suite of tasks.
View details
Scaling Vision with Sparse Mixture of Experts
Carlos Riquelme
Basil Mustafa
Maxim Neumann
Rodolphe Jenatton
André Susano Pinto
Neurips 2021. (2021)
Preview abstract
Sparsely-gated Mixture of Experts networks (MoEs) have demonstrated excellent scalability in Natural Language Processing. In Computer Vision, however, almost all performant networks are "dense", that is, every input is processed by every parameter. We present a Vision MoE (V-MoE), a sparse version of the Vision Transformer, that is scalable and competitive with the largest dense networks. When applied to image recognition, V-MoE matches the performance of state-of-the-art networks, while requiring as little as half of the compute at inference time. Further, we propose an extension to the routing algorithm that can prioritize subsets of each input across the entire batch, leading to adaptive per-image compute. This allows V-MoE to trade-off performance and compute smoothly at test-time. Finally, we demonstrate the potential of V-MoE to scale vision models, and train a 15B parameter model that attains 90.35% on ImageNet.
View details
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Alexey Dosovitskiy
Dirk Weissenborn
Jakob Uszkoreit
Sylvain Gelly
ICLR (2021)
Preview abstract
While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision tasks, attention is usually either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks, while keeping their overall structure in place. We show that this reliance on ConvNets is not necessary and a pure transformer can perform very well on image classification tasks when applied directly to sequences of image patches. When pre-trained on large amounts of data and transferred to multiple recognition benchmarks (ImageNet, CIFAR-10, etc), these transformers attain excellent accuracy, matching or outperforming the best convolutional networks while requiring substantially less computational resources to train.
View details
Self-Supervised Learning of Video-Induced Visual Invariances
Michael Tobias Tschannen
Josip Djolonga
Sylvain Gelly
Conference on Computer Vision and Pattern Recognition (2020)
Preview abstract
We propose a general framework for self-supervised learning of transferable visual representations based on Video-Induced Visual Invariances (VIVI). We make use of the natural hierarchy consisting of (i) frame level invariances (e.g. color and contrast robustness), (ii) shot/clip level invariances (e.g. robustness to changes in object orientation and lighting conditions), and (iii) video level invariances (semantic relationships of scenes across shots/clips) to define a holistic self-supervised loss. We train the proposed model on the YouTube-8M dataset and show that this approach leads to state-of-the-art self-supervised results on the 19 diverse downstream tasks of the Visual Task Adaptation Benchmark (VTAB). We then show how to co-train the model jointly with labeled images, outperforming an ImageNet-pretrained ResNet-50 with $10x$ fewer labeled images.
View details
Underspecification Presents Challenges for Credibility in Modern Machine Learning
Dan Moldovan
Babak Alipanahi
Alex Beutel
Christina Chen
Jon Deaton
Shaobo Hou
Ghassen Jerfel
Yian Ma
Akinori Mitani
Andrea Montanari
Christopher Nielsen
Thomas Osborne
Rajiv Raman
Kim Ramasamy
Martin Gamunu Seneviratne
Shannon Sequeira
Harini Suresh
Victor Veitch
Journal of Machine Learning Research (2020)
Preview abstract
ML models often exhibit unexpectedly poor behavior when they are deployed in real-world domains. We identify underspecification as a key reason for these failures. An ML pipeline is underspecified when it can return many predictors with equivalently strong held-out performance in the training domain. Underspecification is common in modern ML pipelines, such as those based on deep learning. Predictors returned by underspecified pipelines are often treated as equivalent based on their training domain performance, but we show here that such predictors can behave very differently in deployment domains. This ambiguity can lead to instability and poor model behavior in practice, and is a distinct failure mode from previously identified issues arising from structural mismatch between training and deployment domains. We show that this problem appears in a wide variety of practical ML pipelines, using examples from computer vision, medical imaging, natural language processing, clinical risk prediction based on electronic health records, and medical genomics. Our results show the need to explicitly account for underspecification in modeling pipelines that are intended for real-world deployment in any domain.
View details
Big Transfer (BiT): General Visual Representation Learning
Jessica Yung
Sylvain Gelly
ECCV (2020) (to appear)
Preview abstract
Transfer of pre-trained representations improves sample efficiency and simplifies hyperparameter tuning when training deep neural networks for vision. We revisit the paradigm of pre-training on large supervised datasets and fine-tuning the model on a target task. We scale up pre-training, and propose a simple recipe that we call Big Transfer (BiT). By combining a few carefully selected components, and transferring using a simple heuristic, we achieve strong performance on over 20 datasets. BiT performs well across a surprisingly wide range of data regimes -- from 1 example per class to 1M total examples. BiT achieves 87.5% top-1 accuracy on ILSVRC-2012, 99.4% on CIFAR-10, and 76.3% on the 19 task Visual Task Adaptation Benchmark (VTAB). On small datasets, BiT attains 76.8% on ILSVRC-2012 with 10 examples per class, and 97.0% on CIFAR-10 with 10 examples per class. We conduct detailed analysis of the main components that lead to high transfer performance.
View details
On Self-Modulation for Generative Adversarial Networks
Ting Chen
Sylvain Gelly
International Conference on Learning Representations (2019)
Preview abstract
Training Generative Adversarial Networks (GANs) is a notoriously challenging task.
In this work we propose and study an architectural modification, deemed self-modulation, which improves GAN performance across different data sets, architectures, losses, regularizers, and hyperparameter settings. Intuitively, with self-modulation one allows the intermediate feature maps to change as a function of the input z. While reminiscent of other conditioning techniques, it requires no labeled data. Nevertheless, this simple yet effective approach can be readily applied in the conditional setting if side information is available. In a large-scale empirical study we observe a relative decrease of 5%-35% in FID. Furthermore, everything else being equal, just adding this modification to the generator leads improved performance in ~86% of the studied settings which suggest that it can be applied without extensive hyperparameter optimization.
View details
Parameter Efficient Transfer Learning for NLP
Andrei Giurgiu
Stanisław Kamil Jastrzębski
Bruna Halila Morrone
Mona Attariyan
Sylvain Gelly
ICML (2019)
Preview abstract
Fine-tuning large pretrained models is an effective transfer mechanism in NLP. However, in the presence of many downstream tasks, fine-tuning is parameter inefficient: an entire new model is required for every task. As an alternative, we propose transfer with adapter modules. Adapter modules yield a compact and extensible model; they add only a few trainable parameters per task, and new tasks can be added without revisiting previous ones. The parameters of the original network remain fixed, yielding a high degree of parameter sharing. To demonstrate adapter's effectiveness, we transfer the recently proposed BERT Transformer model to 26 diverse text classification tasks, including the GLUE benchmark. Adapters attain near state-of-the-art performance, whilst adding only a few parameters per task. On GLUE, we attain within 0.8% of the performance of full fine-tuning, adding only 3.6% parameters per task. By contrast, fine-tuning trains 100% of the parameters per task.
View details
Preview abstract
We reduce the computational cost of Neural AutoML with transfer learning. AutoML
relieves human effort by automating the design of ML algorithms. Neural
AutoML has become popular for the design of deep learning architectures, however,
this method has a high computation cost.To address this we propose Transfer
Neural AutoML that uses knowledge from prior tasks to speed up network design.
We extend RL-based architecture search methods to support parallel training on
multiple tasks and then transfer the search strategy to new tasks. On language and
image classification data, Transfer Neural AutoML reduces convergence time over
single-task training by over an order of magnitude on many tasks.
View details
Self-Supervised GANs via Auxiliary Rotation Loss
Ting Chen
Conference on Computer Vision and Pattern Recognition (2018)
Preview abstract
Conditional GANs are at the forefront of natural image synthesis. The main drawback of such models is the necessity for labelled data. In this work we exploit two popular unsupervised learning techniques, adversarial training and self-supervision, to close the gap between conditional and unconditional GANs. In particular, we allow the networks to collaborate on the task of representation learning, while being adversarial with respect to the classic GAN game. The role of self-supervision is to encourage the discriminator to learn meaningful feature representations which are not forgotten during training. We test empirically both the quality of the learned image representations, and the quality of the synthesized images. Under the same conditions, the self-supervised GAN attains a similar performance to state-of-the-art conditional counterparts. Finally, we show that this approach to fully unsupervised learning can be scaled to attain an FID of 33 on unconditional ImageNet generation.
View details
Preview abstract
GANs involve training two networks in an adversarial game, where each network's task depends on its adversary. Recently, several works have framed GAN training as an online or continual learning problem. We focus on the discriminator, which must perform classification under an (adversarially) shifting data distribution. When trained on sequential tasks, neural networks exhibit \emph{forgetting}. For GANs, discriminator forgetting leads to training instability. To counter forgetting, we encourage the discriminator to maintain useful representations by adding a self-supervision. Conditional GANs have a similar effect using labels. However, our self-supervised GAN does not require labels, and closes the performance gap between conditional and unconditional models. We show that, in doing so, the self-supervised discriminator learns better representations than regular GANs.
View details
Ask the Right Questions: Active Question Reformulation with Reinforcement Learning
Wei Wang
Sixth International Conference on Learning Representations (2018)
Preview abstract
We frame Question Answering (QA) as a Reinforcement Learning task, an approach that we call Active Question Answering. We propose an agent that sits between the user and a black box QA system and learns to reformulate questions to elicit the best possible answers. The agent probes the system with, potentially many, natural language reformulations of an initial question and aggregates the returned evidence to yield the best answer. The reformulation system is trained end-to-end to maximize answer quality using policy gradient. We evaluate on SearchQA, a dataset of complex questions extracted from Jeopardy!. The agent outperforms a state-of-the-art base model, playing the role of the environment, and other benchmarks. We also analyze the language that the agent has learned while interacting with the question answering system. We find that successful question reformulations look quite different from natural language paraphrases. The agent is able to discover non-trivial reformulation strategies that resemble classic information retrieval techniques such as term re-weighting (tf-idf) and stemming.
View details
Analyzing Language Learned by an Active Question Answering Agent
Wei Wang
Emergent Communication Workshop @ NIPS (2017)
Preview abstract
We analyze the language learned by an agent trained with reinforcement learning as a component of the ActiveQA system [Buck et al., 2017]. In ActiveQA, question answering is framed as a reinforcement learning task in which an agent sits between the user and a black box question-answering system. The agent learns to reformulate the user's questions to elicit the optimal answers. It probes the system with many versions of a question that are generated via a sequence-to-sequence question reformulation model, then aggregates the returned evidence to find the best answer. This process is an instance of machine-machine communication. The question reformulation model must adapt its language to increase the quality of the answers returned, matching the language of the question answering system. We find that the agent does not learn transformations that align with semantic intuitions but discovers through learning classical information retrieval techniques such as tf-idf re-weighting and stemming.
View details
A Scalable Gibbs Sampler for Probabilistic Entity Linking
Advances in Information Retrieval (ECIR 2014), Springer International Publishing, pp. 335-346
Preview abstract
Entity linking involves labeling phrases in text with their referent entities, such as Wikipedia or Freebase entries. This task is challenging due to the large number of possible entities, in the millions, and heavy-tailed mention ambiguity. We formulate the problem in terms of probabilistic inference within a topic model, where each topic is associated with a Wikipedia article. To deal with the large number of topics we propose a novel efficient Gibbs sampling scheme which can also incorporate side information, such as the Wikipedia graph. This conceptually simple probabilistic approach achieves state-of-the-art performance in entity-linking on the Aida-CoNLL dataset.
View details
Stochastic Inference for Scalable Probabilistic Modeling of Binary Matrices
José Miguel Hernández-Lobato
Zoubin Ghahramani
International Conference on Machine Learning (ICML) (2014)
Cold-start Active Learning with Robust Ordinal Matrix Factorization
José Miguel Hernández-Lobato
Zoubin Ghahramani
International Conference on Machine Learning (ICML) (2014)
A Filtering Approach to Stochastic Variational Inference
Probabilistic Matrix Factorization with Non-random Missing Data
José Miguel Hernández-Lobato
Zoubin Ghahramani
International Conference on Machine Learning (ICML) (2014)
Active learning for Interactive Visualization
Tomoharu Iwata
Zoubin Ghahramani
International Conference on Artificial Intelligence and Statistics (AISTATS) (2013)
Statistical Fitting of Undrained Strength Data
Experimental Adaptive Bayesian tomography
Konstantin Kravtsov
Stanislav Straupe
Igor Radchenko
Gleb Struchalin
Sergey Kulik
Physical Review A (2013)
Cognitive Tomography Reveals Complex Task-Independent Mental Representations
Ferenc Huszár
Mohammad Ghassemi
Gergő Orbán
Daniel Wolpert
Máté Lengyel
Current Biology (2013)
Collaborative Gaussian Processes for Preference Learning
Ferenc Huszár
Jose M. Hernández-lobato
Zoubin Ghahramani
Neural Information Processing Systems (NIPS) (2012)
Adaptive Bayesian Quantum Tomography
Bayesian Active Learning for Gaussian Process Classification
Ferenc Huszár
Zoubin Ghahramani
Máté Lengyel
NIPS Workshop on Bayesian optimization, experimental design and bandits: Theory and applications (2011)
