Hierarchical Semantic Loss and Confidence Estimator for Visual-Semantic Embedding-Based Zero-Shot Learning

Traditional supervised learning is dependent on the label of the training data, so there is a limitation that the class label which is not included in the training data cannot be recognized properly. Therefore, zero-shot learning, which can recognize unseen-classes that are not used in training, is gaining research interest. One approach to zero-shot learning is to embed visual data such as images and rich semantic data related to text labels of visual data into a common vector space to perform zero-shot cross-modal retrieval on newly input unseen-class data. This paper proposes a hierarchical semantic loss and confidence estimator to more efficiently perform zero-shot learning on visual data. Hierarchical semantic loss improves learning efficiency by using hierarchical knowledge in selecting a negative sample of triplet loss, and the confidence estimator estimates the confidence score to determine whether it is seen-class or unseen-class. These methodologies improve the performance of zero-shot learning by adjusting distances from a semantic vector to visual vector when performing zero-shot cross-modal retrieval. Experimental results show that the proposed method can improve the performance of zero-shot learning in terms of hit@k accuracy.

Download Full-text

Multi-Task Self-Supervised Learning for Disfluency Detection

Proceedings of the AAAI Conference on Artificial Intelligence ◽

10.1609/aaai.v34i05.6456 ◽

2020 ◽

Vol 34 (05) ◽

pp. 9193-9200

Author(s):

Shaolei Wang ◽

Wangxiang Che ◽

Qi Liu ◽

Pengda Qin ◽

Ting Liu ◽

...

Keyword(s):

Supervised Learning ◽

Large Scale ◽

Experimental Results ◽

Training Data ◽

Competitive Performance ◽

Test Set ◽

Full Dataset ◽

Sentence Classification ◽

Trained Network

Most existing approaches to disfluency detection heavily rely on human-annotated data, which is expensive to obtain in practice. To tackle the training data bottleneck, we investigate methods for combining multiple self-supervised tasks-i.e., supervised tasks where data can be collected without manual labeling. First, we construct large-scale pseudo training data by randomly adding or deleting words from unlabeled news data, and propose two self-supervised pre-training tasks: (i) tagging task to detect the added noisy words. (ii) sentence classification to distinguish original sentences from grammatically-incorrect sentences. We then combine these two tasks to jointly train a network. The pre-trained network is then fine-tuned using human-annotated disfluency detection training data. Experimental results on the commonly used English Switchboard test set show that our approach can achieve competitive performance compared to the previous systems (trained using the full dataset) by using less than 1% (1000 sentences) of the training data. Our method trained on the full dataset significantly outperforms previous methods, reducing the error by 21% on English Switchboard.

Download Full-text

Improving Semi-Supervised Learning for Audio Classification with FixMatch

Electronics ◽

10.3390/electronics10151807 ◽

2021 ◽

Vol 10 (15) ◽

pp. 1807

Author(s):

Sascha Grollmisch ◽

Estefanía Cano

Keyword(s):

Neural Networks ◽

Supervised Learning ◽

Transfer Learning ◽

Data Transfer ◽

State Of The Art ◽

Training Data ◽

Audio Classification ◽

Image Domain ◽

Full Dataset ◽

Audio Data

Including unlabeled data in the training process of neural networks using Semi-Supervised Learning (SSL) has shown impressive results in the image domain, where state-of-the-art results were obtained with only a fraction of the labeled data. The commonality between recent SSL methods is that they strongly rely on the augmentation of unannotated data. This is vastly unexplored for audio data. In this work, SSL using the state-of-the-art FixMatch approach is evaluated on three audio classification tasks, including music, industrial sounds, and acoustic scenes. The performance of FixMatch is compared to Convolutional Neural Networks (CNN) trained from scratch, Transfer Learning, and SSL using the Mean Teacher approach. Additionally, a simple yet effective approach for selecting suitable augmentation methods for FixMatch is introduced. FixMatch with the proposed modifications always outperformed Mean Teacher and the CNNs trained from scratch. For the industrial sounds and music datasets, the CNN baseline performance using the full dataset was reached with less than 5% of the initial training data, demonstrating the potential of recent SSL methods for audio data. Transfer Learning outperformed FixMatch only for the most challenging dataset from acoustic scene classification, showing that there is still room for improvement.

Download Full-text

Experiments of Image Classification Using Dissimilarity Spaces Built with Siamese Networks

Sensors ◽

10.3390/s21051573 ◽

2021 ◽

Vol 21 (5) ◽

pp. 1573

Author(s):

Loris Nanni ◽

Giovanni Minchio ◽

Sheryl Brahnam ◽

Gianluca Maguolo ◽

Alessandra Lumini

Keyword(s):

Vector Space ◽

Image Classification ◽

Ad Hoc ◽

Feature Space ◽

Medical Data ◽

Training Data ◽

Data Sets ◽

Large Set ◽

Clustering Methods ◽

Siamese Networks

Traditionally, classifiers are trained to predict patterns within a feature space. The image classification system presented here trains classifiers to predict patterns within a vector space by combining the dissimilarity spaces generated by a large set of Siamese Neural Networks (SNNs). A set of centroids from the patterns in the training data sets is calculated with supervised k-means clustering. The centroids are used to generate the dissimilarity space via the Siamese networks. The vector space descriptors are extracted by projecting patterns onto the similarity spaces, and SVMs classify an image by its dissimilarity vector. The versatility of the proposed approach in image classification is demonstrated by evaluating the system on different types of images across two domains: two medical data sets and two animal audio data sets with vocalizations represented as images (spectrograms). Results show that the proposed system’s performance competes competitively against the best-performing methods in the literature, obtaining state-of-the-art performance on one of the medical data sets, and does so without ad-hoc optimization of the clustering methods on the tested data sets.

Download Full-text

Laplacian networks: bounding indicator function smoothness for neural networks robustness

APSIPA Transactions on Signal and Information Processing ◽

10.1017/atsip.2021.2 ◽

2021 ◽

Vol 10 ◽

Author(s):

Carlos Lassance ◽

Vincent Gripon ◽

Antonio Ortega

Keyword(s):

Neural Networks ◽

Deep Learning ◽

Supervised Learning ◽

Indicator Function ◽

Training Data ◽

Theoretical Justification ◽

The Past ◽

Noisy Examples

For the past few years, deep learning (DL) robustness (i.e. the ability to maintain the same decision when inputs are subject to perturbations) has become a question of paramount importance, in particular in settings where misclassification can have dramatic consequences. To address this question, authors have proposed different approaches, such as adding regularizers or training using noisy examples. In this paper we introduce a regularizer based on the Laplacian of similarity graphs obtained from the representation of training data at each layer of the DL architecture. This regularizer penalizes large changes (across consecutive layers in the architecture) in the distance between examples of different classes, and as such enforces smooth variations of the class boundaries. We provide theoretical justification for this regularizer and demonstrate its effectiveness to improve robustness on classical supervised learning vision datasets for various types of perturbations. We also show it can be combined with existing methods to increase overall robustness.

Download Full-text

Semi-supervised Intrusive Appliance Load Monitoring in Smart Energy Monitoring System

ACM Transactions on Sensor Networks ◽

10.1145/3448415 ◽

2021 ◽

Vol 17 (3) ◽

pp. 1-20

Author(s):

Vanh Khuyen Nguyen ◽

Wei Emma Zhang ◽

Adnan Mahmood

Keyword(s):

Supervised Learning ◽

Monitoring System ◽

End Users ◽

Training Data ◽

Smart Device ◽

Energy Monitoring ◽

Load Monitoring ◽

Benchmark Datasets ◽

Smart Home Environment ◽

Smart Appliance

Intrusive Load Monitoring (ILM) is a method to measure and collect the energy consumption data of individual appliances via smart plugs or smart sockets. A major challenge of ILM is automatic appliance identification, in which the system is able to determine automatically a label of the active appliance connected to the smart device. Existing ILM techniques depend on labels input by end-users and are usually under the supervised learning scheme. However, in reality, end-users labeling is laboriously rendering insufficient training data to fit the supervised learning models. In this work, we propose a semi-supervised learning (SSL) method that leverages rich signals from the unlabeled dataset and jointly learns the classification loss for the labeled dataset and the consistency training loss for unlabeled dataset. The samples fit into consistency learning are generated by a transformation that is built upon weighted versions of DTW Barycenter Averaging algorithm. The work is inspired by two recent advanced works in SSL in computer vision and combines the advantages of the two. We evaluate our method on the dataset collected from our developed Internet-of-Things based energy monitoring system in a smart home environment. We also examine the method’s performances on 10 benchmark datasets. As a result, the proposed method outperforms other methods on our smart appliance datasets and most of the benchmarks datasets, while it shows competitive results on the rest datasets.

Download Full-text

A Methodology for Utilizing Vector Space to Improve the Performance of a Dog Face Identification Model

Applied Sciences ◽

10.3390/app11052074 ◽

2021 ◽

Vol 11 (5) ◽

pp. 2074

Author(s):

Bohan Yoon ◽

Hyeonji So ◽

Jongtae Rhee

Keyword(s):

Vector Space ◽

Object Identification ◽

Face Identification ◽

Learning Method ◽

Open Set ◽

L2 Norm ◽

The Face ◽

Triplet Loss ◽

Animal Faces ◽

Identification Model

Recent improvements in the performance of the human face recognition model have led to the development of relevant products and services. However, research in the similar field of animal face identification has remained relatively limited due to the greater diversity and complexity in shape and the lack of relevant data for animal faces such as dogs. In the face identification model using triplet loss, the length of the embedding vector is normalized by adding an L2-normalization (L2-norm) layer for using cosine-similarity-based learning. As a result, object identification depends only on the angle, and the distribution of the embedding vector is limited to the surface of a sphere with a radius of 1. This study proposes training the model from which the L2-norm layer is removed by using the triplet loss to utilize a wide vector space beyond the surface of a sphere with a radius of 1, for which a novel loss function and its two-stage learning method. The proposed method classifies the embedding vector within a space rather than on the surface, and the model’s performance is also increased. The accuracy, one-shot identification performance, and distribution of the embedding vectors are compared between the existing learning method and the proposed learning method for verification. The verification was conducted using an open-set. The resulting accuracy of 97.33% for the proposed learning method is approximately 4% greater than that of the existing learning method.

Download Full-text

Deep Learning of Appearance Affinity for Multi-Object Tracking and Re-Identification: A Comparative View

Electronics ◽

10.3390/electronics9111757 ◽

2020 ◽

Vol 9 (11) ◽

pp. 1757

Author(s):

María J. Gómez-Silva ◽

Arturo de la Escalera ◽

José M. Armingol

Keyword(s):

Deep Learning ◽

Object Tracking ◽

Loss Function ◽

Neural Model ◽

Training Data ◽

Learning Approaches ◽

The Core ◽

Triplet Loss ◽

Affinity Model

Recognizing the identity of a query individual in a surveillance sequence is the core of Multi-Object Tracking (MOT) and Re-Identification (Re-Id) algorithms. Both tasks can be addressed by measuring the appearance affinity between people observations with a deep neural model. Nevertheless, the differences in their specifications and, consequently, in the characteristics and constraints of the available training data for each one of these tasks, arise from the necessity of employing different learning approaches to attain each one of them. This article offers a comparative view of the Double-Margin-Contrastive and the Triplet loss function, and analyzes the benefits and drawbacks of applying each one of them to learn an Appearance Affinity model for Tracking and Re-Identification. A batch of experiments have been conducted, and their results support the hypothesis concluded from the presented study: Triplet loss function is more effective than the Contrastive one when an Re-Id model is learnt, and, conversely, in the MOT domain, the Contrastive loss can better discriminate between pairs of images rendering the same person or not.

Download Full-text

A review: preprocessing techniques and data augmentation for sentiment analysis

Computational Social Networks ◽

10.1186/s40649-020-00080-x ◽

2021 ◽

Vol 8 (1) ◽

Author(s):

Huu-Thanh Duong ◽

Tram-Anh Nguyen-Thi

Keyword(s):

Machine Learning ◽

Sentiment Analysis ◽

Supervised Learning ◽

Data Augmentation ◽

Original Data ◽

Training Data ◽

Unseen Data ◽

Augmentation Techniques ◽

User Intervention

AbstractIn literature, the machine learning-based studies of sentiment analysis are usually supervised learning which must have pre-labeled datasets to be large enough in certain domains. Obviously, this task is tedious, expensive and time-consuming to build, and hard to handle unseen data. This paper has approached semi-supervised learning for Vietnamese sentiment analysis which has limited datasets. We have summarized many preprocessing techniques which were performed to clean and normalize data, negation handling, intensification handling to improve the performances. Moreover, data augmentation techniques, which generate new data from the original data to enrich training data without user intervention, have also been presented. In experiments, we have performed various aspects and obtained competitive results which may motivate the next propositions.

Download Full-text

Semi-Supervised Classification and its Application to Filtering IDS False Positives

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.427-429.2309 ◽

2013 ◽

Vol 427-429 ◽

pp. 2309-2312

Author(s):

Hai Bin Mei ◽

Ming Hua Zhang

Keyword(s):

Supervised Learning ◽

Supervised Classification ◽

Classification Performance ◽

False Positives ◽

Training Data ◽

Classification Model ◽

Classification Technique

Alert classifiers built with the supervised classification technique require large amounts of labeled training alerts. Preparing for such training data is very difficult and expensive. Thus accuracy and feasibility of current classifiers are greatly restricted. This paper employs semi-supervised learning to build alert classification model to reduce the number of needed labeled training alerts. Alert context properties are also introduced to improve the classification performance. Experiments have demonstrated the accuracy and feasibility of our approach.

Download Full-text

The Performance of Boolean Retrieval and Vector Space Model in Textual Information Retrieval

CommIT (Communication and Information Technology) Journal ◽

10.21512/commit.v11i1.2108 ◽

2017 ◽

Vol 11 (1) ◽

pp. 33 ◽

Cited By ~ 1

Author(s):

Budi Yulianto ◽

Widodo Budiharto ◽

Iman Herwidiana Kartowisastro

Keyword(s):

Information Retrieval ◽

Vector Space ◽

Vector Space Model ◽

Experimental Results ◽

Inverted Index ◽

Exact Results ◽

Textual Information ◽

Space Model ◽

Corpus Data

Boolean Retrieval (BR) and Vector Space Model (VSM) are very popular methods in information retrieval for creating an inverted index and querying terms. BR method searches the exact results of the textual information retrieval without ranking the results. VSM method searches and ranks the results. This study empirically compares the two methods. The research utilizes a sample of the corpus data obtained from Reuters. The experimental results show that the required times to produce an inverted index by the two methods are nearly the same. However, a difference exists on the querying index. The results also show that the numberof generated indexes, the sizes of the generated files, and the duration of reading and searching an index are proportional with the file number in the corpus and thefile size.

Download Full-text