Learning Deep Latent Spaces for Multi-Label Classification

By Chih-Kuan Yeh, Wei-Chieh Wu, Wei-Jen Ko, Yu-Chiang Frank Wang. 2017

145 citations recorded by the source corpus.

Multi-label classification is a practical yet challenging task in machine learning related fields, since it requires the prediction of more than one label category for each input instance. We propose a novel deep neural networks (DNN) based model, Canonical Correlated AutoEncoder (C2AE), for solving this task. Aiming at better relating feature and label domain data for improved classification, we uniquely perform joint feature and label embedding by deriving a deep latent space, followed by the introduction of label-correlation sensitive loss function for recovering the predicted label outputs. Our C2AE is achieved by integrating the DNN architectures of canonical correlation analysis and autoencoder, which allows end-to-end learning and prediction with the ability to exploit label dependency. Moreover, our C2AE can be easily extended to address the learning problem with missing labels. Our experiments on multiple datasets with different scales confirm the effectiveness and robustness of our proposed method, which is shown to perform favorably against state-of-the-art methods for multi-label classification.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning, Signal Processing

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 Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

In Signal Processing

Citations in the corpus

Elucidating image-to-set prediction: An analysis of models, losses and datasets (2019, background): “Other solutions to capture label co-occurrences include learning joint input and label embeddings with ranking-based losses [53, 26, 56, 24, 12] as well as designing loss functions such as target distribution mean squared error [50] or target distribution cross-entropy [12, 32, 41], which directly a”
Adversarial Multi-label Prediction for Spoken and Visual Signal Tagging (2019, methodology): “We follow the same protocol in [11] to randomly select r data points in each dataset for training/validation and the rest are for testing.”
Multi-label Zero-Shot Learning with Structured Knowledge Graphs (2017, methodology): “Label embedding methods are among the popular techniques, which transform labels into embedded label vectors, so that the correlation between labels can be exploited [47, 43, 18, 29, 49].”
The Emerging Trends of Multi-Label Learning (2020, methodology): “C2AE consists of two DNN modules, i.e., DCCA and autoencoder, and seeks three mapping functions, i.e., feature mapping Fx, encoding function Fe, and decoding function Fd.”
CU-Net: Component Unmixing Network for Textile Fiber Identification (2019, methodology): “We compared our methods (CU-Net, CU-Net SI RI , CU-Net RI and CU-Net SI ) with SB-MLC-Net and the state-of-the-art methods (Wang et al. 2016; Yeh et al. 2017).”
Inverse Cooking: Recipe Generation From Food Images (2018, background): “Moreover, joint input and label embeddings [57, 25, 61] have been introduced to preserve correlations and predict label sets.”
Multi-Label Image Classification via Knowledge Distillation from Weakly-Supervised Detection (2018, methodology): “performance and efficiency. 2 RELATED WORK Multi-LabelImageClassification(MLIC).TheprogressofMLIC [9, 28, 34, 45] has been greatly made with deep convolutional neural network [11, 16, 25]. Some works [13, 15, 18, 19, 21, 27, 32, 37, 38, 41, 44] embed label dependencies with the deep model to improve”
Deep Streaming Label Learning (2020, methodology): “Exploring and exploiting the relationship between labels has been demonstrated to be advantageous and crucial for boosting MLL performance (Zhang & Zhou, 2014; Yeh et al., 2017; Menon et al., 2019).”
CIFDM: Continual and Interactive Feature Distillation for Multi-Label Stream Learning (2021, methodology): “For example, C2AE [23] applies the DNN model as feature encoder to utilize features and label embedding jointly; SLEEC[4] learns a set of embeddings to preserve pairwise distances between only the nearest label vectors, and accurately predict infrequently occurring (tail) labels.”
Multi-Label Learning with Global and Local Label Correlation (2017, background): “Because the labels are correlated in multi-label learning, the label matrix is often assumed to be low-rank [22], [24].”
Order-Free RNN with Visual Attention for Multi-Label Classification (2017, background): “2016) learn label dependencies with graphical structure, and latent space methods (Yeh et al. 2017; Bhatia et al. 2015) choose to project features and labels into a common latent space.”
Neural Message Passing for Multi-Label Classification (2019, background): “The second group learns a shared latent space representing both input features and output labels, and then upsamples from the space to reconstruct the target labels [33, 5].”