Set-valued classification -- overview via a unified framework

Chzhen, Evgenii; Denis, Christophe; Hebiri, Mohamed; Lorieul, Titouan

doi:10.48550/arxiv.2102.12318

Cited by 5 publications

(6 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, a slight adaptation of our loss for imbalanced datasets (leveraging uneven margins) outperforms other baseline losses. Studying deep learning optimization methods for other set-valued classification tasks, such as average size control or point-wise error control (Chzhen et al, 2021) are left for future work.…”

Section: Discussionmentioning

confidence: 99%

Stochastic smoothing of the top-K calibrated hinge loss for deep imbalanced classification

Garcin¹,

Servajean²,

Joly³

et al. 2022

Preprint

View full text Add to dashboard Cite

In modern classification tasks, the number of labels is getting larger and larger, as is the size of the datasets encountered in practice. As the number of classes increases, class ambiguity and class imbalance become more and more problematic to achieve high top-1 accuracy. Meanwhile, Top-K metrics (metrics allowing K guesses) have become popular, especially for performance reporting. Yet, proposing top-K losses tailored for deep learning remains a challenge, both theoretically and practically. In this paper we introduce a stochastic top-K hinge loss inspired by recent developments on top-K calibrated losses. Our proposal is based on the smoothing of the top-K operator building on the flexible "perturbed optimizer" framework. We show that our loss function performs very well in the case of balanced datasets, while benefiting from a significantly lower computational time than the state-of-the-art top-K loss function. In addition, we propose a simple variant of our loss for the imbalanced case. Experiments on a heavy-tailed dataset show that our loss function significantly outperforms other baseline loss functions.

show abstract

Section: Discussionmentioning

confidence: 99%

Stochastic smoothing of the top-K calibrated hinge loss for deep imbalanced classification

Garcin¹,

Servajean²,

Joly³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Set-valued classification is proposed, which infers a set of up to candidates instead of predicting only one correct answer as in ordinary classification [8,25]. There have been works on objective functions with good classification accuracy while maintaining a small prediction-set size [26]. As set-valued classification consists of a set of classes, a separate verification is required to make a final decision of class.…”

Section: Overlapping and Imbalanced Distributionsmentioning

confidence: 99%

Deep Attention-based Classification of Plant Images with Hierarchical Similarity and Imbalanced Distribution

Kim¹,

Kim²

2023

IJETAE

View full text Add to dashboard Cite

—Standard datasets in deep learning-basedimage classification usually provide two favorable conditions for design: preservation of visual homogeneity of each category and uniform distribution of sample images. These conditions are not assured in a dataset of plant images. Two different species of plants under the same genus look very similar and the number of collectible images has a large variation over species.Visual similarity, however, can be turned into advantage in hierarchical approach, by assigning two-fold labels of species and genus, in two phases ofrough classification of genera and fine classification of species. We propose a hierarchical classification in which the concatenation scheme is augmented with a channel attention which focuses on the sibling relation of species. We compared our method with flat classification and conventional hierarchical classification. The test was on a PlantNet-300K dataset 300k images, composed of 303 genera and 1081 species. In experimental results, the channel attention layers lead to stable discerningof the minute difference among visually similar species. The proposed hierarchical classification method outperforms both the flat classification and the conventional hierarchical classification.

show abstract

“…To evaluate the model, we chose not to use pseudo-absences because of the bias induced by such methods (Phillips et al, 2009;Botella et al, 2020). Instead, we used a set-valued metric (Chzhen et al, 2021) to assess the quality of the species assemblage predicted by the model for a given input. Specifically, we chose the commonly used top-k accuracy as suggested in Botella et al (2019).…”

Section: Evaluation Metricsmentioning

confidence: 99%

Deep Species Distribution Modeling From Sentinel-2 Image Time-Series: A Global Scale Analysis on the Orchid Family

et al. 2022

View full text Add to dashboard Cite

Species distribution models (SDMs) are widely used numerical tools that rely on correlations between geolocated presences (and possibly absences) and environmental predictors to model the ecological preferences of species. Recently, SDMs exploiting deep learning and remote sensing images have emerged and have demonstrated high predictive performance. In particular, it has been shown that one of the key advantages of these models (called deep-SDMs) is their ability to capture the spatial structure of the landscape, unlike prior models. In this paper, we examine whether the temporal dimension of remote sensing images can also be exploited by deep-SDMs. Indeed, satellites such as Sentinel-2 are now providing data with a high temporal revisit, and it is likely that the resulting time-series of images contain relevant information about the seasonal variations of the environment and vegetation. To confirm this hypothesis, we built a substantial and original dataset (called DeepOrchidSeries) aimed at modeling the distribution of orchids on a global scale based on Sentinel-2 image time series. It includes around 1 million occurrences of orchids worldwide, each being paired with a 12-month-long time series of high-resolution images (640 x 640 m RGB+IR patches centered on the geolocated observations). This ambitious dataset enabled us to train several deep-SDMs based on convolutional neural networks (CNNs) whose input was extended to include the temporal dimension. To quantify the contribution of the temporal dimension, we designed a novel interpretability methodology based on temporal permutation tests, temporal sampling, and temporal averaging. We show that the predictive performance of the model is greatly increased by the seasonality information contained in the temporal series. In particular, occurrence-poor species and diversity-rich regions are the ones that benefit the most from this improvement, revealing the importance of habitat's temporal dynamics to characterize species distribution.

show abstract

Set-valued classification -- overview via a unified framework

Cited by 5 publications

References 37 publications

Stochastic smoothing of the top-K calibrated hinge loss for deep imbalanced classification

Stochastic smoothing of the top-K calibrated hinge loss for deep imbalanced classification

Deep Attention-based Classification of Plant Images with Hierarchical Similarity and Imbalanced Distribution

Deep Species Distribution Modeling From Sentinel-2 Image Time-Series: A Global Scale Analysis on the Orchid Family

Contact Info

Product

Resources

About