Deep Bayesian Active Learning with Image Data

By Yarin Gal, Riashat Islam, Zoubin Ghahramani. 2017

2,034 citations recorded by the source corpus.

Even though active learning forms an important pillar of machine learning, deep learning tools are not prevalent within it. Deep learning poses several difficulties when used in an active learning setting. First, active learning (AL) methods generally rely on being able to learn and update models from small amounts of data. Recent advances in deep learning, on the other hand, are notorious for their dependence on large amounts of data. Second, many AL acquisition functions rely on model uncertainty, yet deep learning methods rarely represent such model uncertainty. In this paper we combine recent advances in Bayesian deep learning into the active learning framework in a practical way. We develop an active learning framework for high dimensional data, a task which has been extremely challenging so far, with very sparse existing literature. Taking advantage of specialised models such as Bayesian convolutional neural networks, we demonstrate our active learning techniques with image data, obtaining a significant improvement on existing active learning approaches. We demonstrate this on both the MNIST dataset, as well as for skin cancer diagnosis from lesion images (ISIC2016 task).

Subfields: Artificial Intelligence, Ocean Engineering, Representation & Self-Supervised Learning, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Ocean Engineering

In Representation & Self-Supervised Learning

In Evaluation & Benchmarks

Citations in the corpus

Scalable agent alignment via reward modeling: a research direction (2018, methodology): “While Bayesian methods can help with calibration (Gal et al., 2017a), they are insufﬁcient in practice for deep neural networks (Kuleshov et al., 2018).”
Making deep neural networks right for the right scientific reasons by interacting with their explanations (2020, background): “Prototypical active learners include max-margin [43] and Bayesian approaches [44]; recently, deep variants have been proposed [45].”
Uncertainty in Multitask Transfer Learning (2018, background): “While we no longer have a well calibrated task uncertainty, the KL term still acts as an effective regularizer and prevents overﬁtting on small datasets 12 .”
Bayesian Model-Agnostic Meta-Learning (2018, background): “It brings us not only robustness to overﬁtting but also numerous beneﬁts such as improved prediction accuracy by Bayesian ensembling (Balan et al., 2015), active learning (Gal et al., 2016), and principled/safe exploration in reinforcement learning (Houthooft et al., 2016).”
Bayesian Hypernetworks (2017, methodology): “We consider all of the acquisition functions listed in (Gal et al., 2017) as possible scores for the AOPR and AOROC metrics, but found that the maximum conﬁdence of the softmax probabilities (i.e. "variation ratio") acquisition function used by Hendrycks and Gimpel (2016) gave the best performance.”
Few-Shot Continual Active Learning by a Robot (2022, methodology): “Unlike most active learning setups [18, 14], in FoCAL the model does not have access to a training set in each increment for training the model to predict uncertain object classes.”
Reliable Uncertainty Estimates in Deep Neural Networks using Noise Contrastive Priors (2018, background): “Recently, Gal et al. (2017) applied a convolutional neural network with dropout uncertainty to images.”
Towards Safe Autonomous Driving: Capture Uncertainty in the Deep Neural Network For Lidar 3D Vehicle Detection (2018, methodology): “This dropout sampling method has been applied to active learning for cancer diagnosis [25], semantic segmentation [26], [27] and image object detection for openset problem [28].”
The limits and potentials of deep learning for robotics (2018, methodology): “Some current approaches [24], [29] leverage the uncertainty estimation techniques based on approximate Bayesian inference (see Section II-A.1) to choose the most informative samples.”
A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks (2018, background): “, active learning [8], ensemble learning [19] and few-shot learning [31].”
A Wholistic View of Continual Learning with Deep Neural Networks: Forgotten Lessons and the Bridge to Active and Open World Learning (2020, methodology): “[110].”
Explanatory Interactive Machine Learning (2019, background): “Making deep active learning [12] explanatory is likely to improve upon the sample complexity of deep learning.”
A survey of uncertainty in deep neural networks (2021, methodology): “used in several works [23], [24], [294], [25], [26].”
FITAnnotator: A Flexible and Intelligent Text Annotation System (2021, background): “A framework where a model learns from small amounts of data, and optimizes the selection of the most informative or diverse sample to annotate in order to maximize training utility value, is referred to as active learning (Gal et al., 2017; Schröder and Niekler, 2020).”
The Power of Ensembles for Active Learning in Image Classification (2018, methodology): “AL for CNNs Besides [16] there are three other AL approaches for CNNs to the best of our knowledge: [52] use a density approach to cover the entire space of unlabeled data points using a geometric based similarity function between images.”
A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges (2020, neutral): “DBAL [101], i.”
Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds (2019, methodology): “For instance, Gal et al. (2017) use Dropout as an approximation of the posterior of the model parameters, and develop information-based uncertainty reduction criteria; inspired by recent advances on adversarial examples generation, Ducoffe and Precioso (2018) use the distance between an example and…”