Variational Information Bottleneck for Effective Low-Resource Audio Classification

Si, Shijing; Wang, Jianzong; Sun, Huiming; Wu, Jianhan; Zhang, Chuanyao; Qu, Xiaobo; Cheng, Ning; Chen, Lei; Xiao, Jing

doi:10.21437/interspeech.2021-2028

Cited by 8 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Ever since the adoption of deep neural networks (DNNs), there have been many developments in ASC systems that treat the above problems with end-to-end learning methods. In particular, convolutional neural networks (CNNs) [10][11][12] significantly affect feature learning from pre-processed raw features, such as the log-mel spectrogram, mel-frequency cepstral coefficients (MFCC), constant-Q transform (CQT), gammatone frequency cepstral coefficients (GFCC), and chromagram [13][14][15]. The common strategy for designing an ASC system is to extract high-level feature maps that consist of activation scores of events from the raw audio data and classify them with global average pooling (GAP) or fully connected layers (FC layers) to each scene through supervised learning methods.…”

Section: Introductionmentioning

confidence: 99%

Instance-level loss based multiple-instance learning framework for acoustic scene classification

Choi¹,

Chang²,

Yang³

et al. 2022

Preprint

View full text Add to dashboard Cite

In acoustic scene classification (ASC) task, an acoustic scene consists of diverse attributes and is inferred by identifying combinations of some distinct attributes among them. This study aims to extract and cluster these attributes effectively using a multiple-instance learning (MIL) framework for ASC. MIL, known as one of the weakly supervised learning methods, is a way to extract instances from input data and infer a scene corresponding to the input data with those unlabeled instances. We develop the MIL framework more suitable for ASC systems, adopting instance-level labels and instance-level loss, which are effective in extracting and clustering instances. As a result, the witness rate increases significantly compared to the framework without instance-level loss and labels. Also in several MIL-based ASC systems, the classification accuracy improves by about 5 to 11% than without instance-level loss. In addition, we designed a fully separated convolutional module which is a low-complexity neural network consisting of pointwise, frequency-sided depthwise, and temporal-sided depthwise convolutional filters. Considering both complexity and performance, our proposed system is more practical compared to previous systems on the DCASE 2019 challenge task 1-A leader board. We surpassed the third-place model by achieving a performance of 82.3% with only the model complexity of 417K, which is at least 40 times fewer than other systems.

show abstract

Section: Introductionmentioning

confidence: 99%

Instance-level loss based multiple-instance learning framework for acoustic scene classification

Choi¹,

Chang²,

Yang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…To alleviate overfitting and enhance the generalization ability of deep learning models, many methods have been proposed, such as data augmentation [6], dropout [7], batch normalization [8], weight decay [9], pretraining [10], variational information bottleneck [11], etc. The most recently developed method is R-Drop [12], which forces the output distributions of two different sub-networks generated from dropout by utilizing Kullback-Leibler (KL) divergence [13] in the training stage to be consistent.…”

Section: Introductionmentioning

confidence: 99%

Augmentation-induced Consistency Regularization for Classification

Wu¹,

Si²,

Wang³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Deep neural networks have become popular in many supervised learning tasks, but they may suffer from overfitting when the training dataset is limited. To mitigate this, many researchers use data augmentation, which is a widely used and effective method for increasing the variety of datasets. However, the randomness introduced by data augmentation causes inevitable inconsistency between training and inference, which leads to poor improvement. In this paper, we propose a consistency regularization framework based on data augmentation, called CR-Aug, which forces the output distributions of different sub models generated by data augmentation to be consistent with each other. Specifically, CR-Aug evaluates the discrepancy between the output distributions of two augmented versions of each sample, and it utilizes a stop-gradient operation to minimize the consistency loss. We implement CR-Aug to image and audio classification tasks and conduct extensive experiments to verify its effectiveness in improving the generalization ability of classifiers. Our CR-Aug framework is ready-to-use, it can be easily adapted to many state-of-the-art network architectures. Our empirical results show that CR-Aug outperforms baseline methods by a significant margin.

show abstract

Boosting StarGANs for Voice Conversion with Contrastive Discriminator

Wang²,

Zhang³

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Variational Information Bottleneck for Effective Low-Resource Audio Classification

Cited by 8 publications

References 0 publications

Instance-level loss based multiple-instance learning framework for acoustic scene classification

Instance-level loss based multiple-instance learning framework for acoustic scene classification

Augmentation-induced Consistency Regularization for Classification

Boosting StarGANs for Voice Conversion with Contrastive Discriminator

Contact Info

Product

Resources

About