One-Shot Learning with a Hierarchical Nonparametric Bayesian Model

By Ruslan Salakhutdinov, Joshua B. Tenenbaum, Josh Tenenbaum, Antonio Torralba. 2011

177 citations recorded by the source corpus.

Subfields: Probabilistic & Bayesian Methods, Meta & Continual Learning

How researchers see this paper

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

In Probabilistic & Bayesian Methods

In Meta & Continual Learning

Citations in the corpus

On First-Order Meta-Learning Algorithms (2018, background): “[17, 15] showed that human subjects can recognize new object categories based on a few example images.”
Data Augmentation Generative Adversarial Networks (2017, methodology): “Few-Shot Learning and Meta-Learning: there have been a number of approaches to few shot learning, from (Salakhutdinov et al., 2012) where they use a hierarchical Boltzmann machine, through to modern deep learning architectures for one-shot conditional generation in (Rezende et al., 2016),…”
Recasting Gradient-Based Meta-Learning as Hierarchical Bayes (2018, methodology): “A variety of inference methods have been used in Bayesian models, including exact inference (Lake et al., 2011), sampling meth-ods (Salakhutdinov et al., 2012), and variational methods (Edwards & Storkey, 2017).”
Low-Shot Learning from Imaginary Data (2018, background): “One class of approaches builds generative models that can share priors across categories [7, 25, 10].”
Deep Meta-Learning: Learning to Learn in the Concept Space (2018, methodology): “(Salakhutdinov et al., 2012) organize seen categories into super-categories to derive hierarchically structured priors for new categories using a hierarchical Bayesian model.”
Taskonomy: Disentangling Task Transfer Learning (2018, background): “compositional modeling [33, 8, 11, 21, 53, 89, 87], homomorphic cryptography [40], lifelong learning [90, 13, 82, 81], functional maps [68], certain aspects of Bayesian inference and Dirichlet processes [52, 88, 87, 86, 35, 37], few-shot learning [78, 23, 22, 67, 83], transfer learning [72, 81, 27, ”
Learning Like a Child: Fast Novel Visual Concept Learning from Sentence Descriptions of Images (2015, methodology): “For the computer vision field, several methods are proposed [16, 46, 53, 31] to handle the problem of learning new Figure 1: An illustration of the Novel Visual Concept learning from Sentences (NVCS) task.”
One-Shot Imitation from Observing Humans via Domain-Adaptive Meta-Learning (2018, background): “functions, and the fine-tuning process as inference under the learned prior [42, 17].”
Generalizing from a Few Examples (2019, methodology): “ strategy finds the bestθwhich parameterize h ∈Hfor these super classes as prior knowledge. A new class is first assigned to super class, then finds the besth through adapting the super class’s h. In [97], they learn to form a hierarchy of classes using D c’s (which includes D), whose θis learned us”
Few-Shot Unsupervised Image-to-Image Translation (2019, background): “Early works use generative models of appearance that share priors across classes in a hierarchical manner [10, 36].”
Domain Adaptive Dialog Generation via Meta Learning (2019, background): “, 2018), few-shot learning (Salakhutdinov et al., 2012; Li et al., 2006; Norouzi et al., 2013; Socher et al., 2013) and meta-learning (Finn et al.”
Low-shot visual object recognition (2016, background): “The benefit of such approaches is that they can tap into a global [10] or a supercategorylevel [33] prior.”
Meta-Learning to Detect Rare Objects (2019, background): “However, in many practical applications such as self-driving vehicles, recognition systems need to rapidly recognize some neverbefore-seen objects from a very limited number of examples [16, 49, 3].”
Multi-attention Network for One Shot Learning (2017, methodology): “The method in [19] takes a step further.”
One-Shot Learning for Semantic Segmentation (2017, background): “While the problem is similar to k -shot learning, which has been extensively studied for image classiﬁcation [36, 39], applying it to segmentation requires some modiﬁcation.”
Low-Shot Visual Recognition by Shrinking and Hallucinating Features (2016, background): “The benefit of such approaches is that they can tap into a global [13] or a supercategory-level [36] prior.”