DEEP NEAREST CLASS MEAN CLASSIFIERS

By Samantha Guerriero, Barbara Caputo, Thomas Mensink. 2018

How researchers see this paper

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

Citations in the corpus

Incremental Learning in Online Scenario (2020, methodology): “We apply NCM in the beginning and update using an online estimate of the class mean [7] for each new observation.”
Decoupling Representation and Classifier for Long-Tailed Recognition (2019, methodology): “…is to ﬁrst compute the mean feature representation for each class on the training set and then perform nearest neighbor search either using cosine similarity or the Euclidean distance computed on L 2 normalized mean features (Snell et al., 2017; Guerriero et al., 2018; Rebufﬁ et al., 2017).”
Adding New Tasks to a Single Network with Weight Trasformations using Binary Masks (2018, background): “Recently, various works have addressed these problems within the framework of deep architectures [31, 10, 1].”
Deep Repulsive Prototypes for Adversarial Robustness (2021, methodology): “Instead of constantly re-estimating and calibrating the prototypes – as in Mustafa et al. (2019) or others (Snell, Swersky, and Zemel 2017; Guerriero, Caputo, and Mensink 2018) – they propose to define the output space as a hypersphere and partition it into predefined class prototypes.”
Hyperspherical Prototype Networks (2019, unknown): “We compare to Deep NCM of Guerriero et al. [11], since it can handle any number of training examples and any output dimensionality, akin to our approach.”
Compositional Prototype Network with Multi-view Comparision for Few-Shot Point Cloud Semantic Segmentation (2020, background): “DeepNCM [9] directly learn highly non-linear deep representations of data based on metric learning, and the DeepNCM are proved to alleviate class-imbalance effect in [10].”
Knowledge is Never Enough: Towards Web Aided Deep Open World Recognition (2019, methodology): “We build on recent work by Guerriero et al [6] and present the first deep open world recognition architecture.”