Active Learning for Convolutional Neural Networks: A Core-Set Approach
arXiv:1708.00489
Abstract
Convolutional neural networks (CNNs) have been successfully applied to many recognition and learning tasks using a universal recipe; training a deep model on a very large dataset of supervised examples. However, this approach is rather restrictive in practice since collecting a large set of labeled images is very expensive. One way to ease this problem is coming up with smart ways for choosing images to be labelled from a very large collection (ie. active learning). Our empirical study suggests that many of the active learning heuristics in the literature are not effective when applied to CNNs in batch setting. Inspired by these limitations, we define the problem of active learning as core-set selection, ie. choosing set of points such that a model learned over the selected subset is competitive for the remaining data points. We further present a theoretical result characterizing the performance of any selected subset using the geometry of the datapoints. As an active learning algorithm, we choose the subset which is expected to yield best result according to our characterization. Our experiments show that the proposed method significantly outperforms existing approaches in image classification experiments by a large margin.
ICLR 2018 Paper
References in corpus (3)
Cited by in corpus (63)
- Uncertainty Quantification in Machine Learning for Engineering Design and Health Prognostics: A Tutorial
- Active label cleaning for improved dataset quality under resource constraints
- Active learning for data streams: a survey
- A Deep Active Survival Analysis Approach for Precision Treatment Recommendations: Application of Prostate Cancer
- Decentralized Deep Learning for Multi-Access Edge Computing: A Survey on Communication Efficiency and Trustworthiness
- Computer Vision with Deep Learning for Plant Phenotyping in Agriculture: A Survey
- Soft-Label Anonymous Gastric X-ray Image Distillation
- Weakly Supervised Vessel Segmentation in X-ray Angiograms by Self-Paced Learning from Noisy Labels with Suggestive Annotation
- Mapping oil palm density at country scale: An active learning approach
- Supporting DNN Safety Analysis and Retraining through Heatmap-based Unsupervised Learning
- Efficient Quality Control of Whole Slide Pathology Images with Human-in-the-loop Training
- REB: Reducing Biases in Representation for Industrial Anomaly Detection
- A Survey on Deep Learning of Small Sample in Biomedical Image Analysis
- A non-cooperative meta-modeling game for automated third-party calibrating, validating, and falsifying constitutive laws with parallelized adversarial attacks
- Deep Indexed Active Learning for Matching Heterogeneous Entity Representations
- Federated Active Learning (F-AL): an Efficient Annotation Strategy for Federated Learning
- Deep Active Learning for Computer Vision: Past and Future
- A semi-supervised self-training method to develop assistive intelligence for segmenting multiclass bridge elements from inspection videos
- Graph Policy Network for Transferable Active Learning on Graphs
- Graph Convolutional Networks based on Manifold Learning for Semi-Supervised Image Classification
- Deep Learning on a Data Diet: Finding Important Examples Early in Training
- MUS-CDB: Mixed Uncertainty Sampling with Class Distribution Balancing for Active Annotation in Aerial Object Detection
- Batch Active Preference-Based Learning of Reward Functions
- Towards Active Learning for Action Spotting in Association Football Videos
- ScatterShot: Interactive In-context Example Curation for Text Transformation
- SIMILAR: Submodular Information Measures Based Active Learning In Realistic Scenarios
- Model-Centric and Data-Centric Aspects of Active Learning for Deep Neural Networks
- A Survey on Cost Types, Interaction Schemes, and Annotator Performance Models in Selection Algorithms for Active Learning in Classification
- Efficient Visual Recognition with Deep Neural Networks: A Survey on Recent Advances and New Directions
- Active Learning for Visual Question Answering: An Empirical Study
- A Self Supervised StyleGAN for Image Annotation and Classification with Extremely Limited Labels
- ALGAN: Anomaly Detection by Generating Pseudo Anomalous Data via Latent Variables
- Active Learning for Deep Detection Neural Networks
- Small-Text: Active Learning for Text Classification in Python
- Active Learning to Classify Macromolecular Structures in situ for Less Supervision in Cryo-Electron Tomography
- Minority Class Oriented Active Learning for Imbalanced Datasets
- Exploring Active Learning for Semiconductor Defect Segmentation
- A Mathematical Analysis of Learning Loss for Active Learning in Regression
- Exploring Spatial Diversity for Region-based Active Learning
- JuryGCN: Quantifying Jackknife Uncertainty on Graph Convolutional Networks
- Deep Active Learning for Text Classification with Diverse Interpretations
- Homogeneous Learning: Self-Attention Decentralized Deep Learning
- Exploring UMAP in hybrid models of entropy-based and representativeness sampling for active learning in biomedical segmentation
- Bayesian Estimate of Mean Proper Scores for Diversity-Enhanced Active Learning
- Auxo: Efficient Federated Learning via Scalable Client Clustering
- Forgetful Active Learning with Switch Events: Efficient Sampling for Out-of-Distribution Data
- A Simple yet Effective Framework for Active Learning to Rank
- Towards Balanced Active Learning for Multimodal Classification
- One Line To Rule Them All: Generating LO-Shot Soft-Label Prototypes
- Collaborative Intelligence Orchestration: Inconsistency-Based Fusion of Semi-Supervised Learning and Active Learning
- High-contrast "gaudy" images improve the training of deep neural network models of visual cortex
- Region and Spatial Aware Anomaly Detection for Fundus Images
- Feature Space Exploration For Planning Initial Benthic AUV Surveys
- On the Robustness of Active Learning
- Marich: A Query-efficient Distributionally Equivalent Model Extraction Attack using Public Data
- Cost-Based Budget Active Learning for Deep Learning
- One-bit Supervision for Image Classification: Problem, Solution, and Beyond
- Analyzing Data Selection Techniques with Tools from the Theory of Information Losses
- Bayesian Uncertainty and Expected Gradient Length -- Regression: Two Sides Of The Same Coin?
- Confidence-Aware Active Feedback for Interactive Instance Search
- Comparative study of subset selection methods for rapid prototyping of 3D object detection algorithms
- Observations on K-image Expansion of Image-Mixing Augmentation for Classification
- Loss Prediction: End-to-End Active Learning Approach For Speech Recognition