Task2Vec: Task Embedding for Meta-Learning

Ontology learning from text is one of the knowledge acquisition processes that facilitates construction of ontology. Considerable research is being done on learning concepts and relations, especially on acquiring semantic relations between concepts for a specific domain. However, more of the research contributions are in learning either taxonomic relations or semantic relations but not in both. Even those few research works that address learning of both types of relations deal with simple sentences only resulting in low recall value. Further, these approaches are semi-automatic, which require either user's feedback or domain expert's knowledge. In this paper, we propose a single framework that is automatic and domain-independent that helps in learning both taxonomic and non-taxonomic relations. We have developed a clause-based approach that automatically extracts the relations for concepts from unstructured text documents. Our approach is capable of handling complex sentences by identifying hidden triples present in the sentences. We have evaluated our methodology of relation learning for the concepts specified by AGROVOC and Open Directory Project using a corpus of web documents. The precision, recall and F1-measure of our method were observed to be considerably higher than those of state-of-the-art methodologies for relation learning.

Cooperative image analysis in visual sensor networks

The feature extraction is the process to represent raw image in a reduced form to facilitate decision making such as pattern detection, classification or recognition. Finding and extracting reliable and discriminative features is always a crucial step to complete the task of image recognition and computer vision. Furthermore, as the number of application demands increase, an extended study and investigation in the feature extraction field becomes very important. The goal of this chapter is to present an intensive survey of existing literatures on feature extraction techniques over the last years. All these techniques and algorithms have their advantages and limitations. Thus, in this chapter analysis of various techniques and transformations, submitted earlier in literature, for extracting various features from images will be discussed. Additionally, future research directions in the feature extraction area are provided.

Objectives: The research work aims to develop an automated detection system that uses video captures to identify roadside accidents or significant events. By alerting nearby hospitals and emergency services, the system reduces response times and potentially saves lives. The system’s integration with existing emergency response systems ensures prompt assistance to those in need. Methods: The video processing pipeline begins by converting video files into frames for analysis and feature extraction. These features serve as inputs for classification algorithms such as Random Forest, SVM, and KNN. The model’s performance is evaluated using a training set and unseen test data, with the predicted classifications compared against the ground truth labels. Findings: Among the tested classification algorithms, the Random Forest algorithm achieved the highest accuracy. Using 128 frames for analysis provided more comprehensive information, yielding a 96% accuracy rate. This combination proves to be a powerful tool in classification tasks, providing reliable and accurate outputs. Novelty: Machine learning algorithms are instrumental in automating accident detection from video captures. They analyse video footage to identify accidents and promptly alert relevant authorities. This technology can also dispatch emergency messages to nearby hospitals, ensuring quick assistance. The consideration of different frame counts in classification improves accuracy by capturing critical moments and patterns. Machine learning algorithms applied in this work significantly enhance emergency response, reduce response times, and potentially save lives. Keywords: ­ Emergency Services, Accidents, Feature Extraction, Random Forest, SVM, KNN

The choice of the loss function is critical in extreme multi-label learning where the objective is to annotate each data point with the most relevant subset of labels from an extremely large label set. Unfortunately, existing loss functions, such as the Hamming loss, are unsuitable for learning, model selection, hyperparameter tuning and performance evaluation. This paper addresses the issue by developing propensity scored losses which: (a) prioritize predicting the few relevant labels over the large number of irrelevant ones; (b) do not erroneously treat missing labels as irrelevant but instead provide unbiased estimates of the true loss function even when ground truth labels go missing under arbitrary probabilistic label noise models; and (c) promote the accurate prediction of infrequently occurring, hard to predict, but rewarding tail labels. Another contribution is the development of algorithms which efficiently scale to extremely large datasets with up to 9 million labels, 70 million points and 2 million dimensions and which give significant improvements over the state-of-the-art. This paper's results also apply to tagging, recommendation and ranking which are the motivating applications for extreme multi-label learning. They generalize previous attempts at deriving unbiased losses under the restrictive assumption that labels go missing uniformly at random from the ground truth. Furthermore, they provide a sound theoretical justification for popular label weighting heuristics used to recommend rare items. Finally, they demonstrate that the proposed contributions align with real world applications by achieving superior clickthrough rates on sponsored search advertising in Bing.

Visual illustration transformation from real-world to cartoon images is one of the famous and challenging tasks in computer vision. Image-to-image translation from real-world to cartoon domains poses issues such as a lack of paired training samples, lack of good image translation, low feature extraction from the previous domain images, and lack of high-quality image translation from the traditional generator algorithms. To solve the above-mentioned issues, paired independent model, high-quality dataset, Bayesian-based feature extractor, and an improved generator must be proposed. In this study, we propose a high-quality dataset to reduce the effect of paired training samples on the model’s performance. We use a Bayesian Very Deep Convolutional Network (VGG)-based feature extractor to improve the performance of the standard feature extractor because Bayesian inference regularizes weights well. The generator from the Cartoon Generative Adversarial Network (GAN) is modified by introducing a depthwise convolution layer and channel attention mechanism to improve the performance of the original generator. We have used the Fréchet inception distance (FID) score and user preference score to evaluate the performance of the model. The FID scores obtained for the generated cartoon and real-world images are 107 and 76 for the TCC style, and 137 and 57 for the Hayao style, respectively. User preference score is also calculated to evaluate the quality of generated images and our proposed model acquired a high preference score compared to other models. We achieved stunning results in producing high-quality cartoon images, demonstrating the proposed model’s effectiveness in transferring style between authentic images and cartoon images.

A key task in computer vision is image retrieval, which has wide-ranging applications across multiple domains. Using query picture features, this abstract proposes an image retrieval approach that focuses on the widely used Scale-Invariant Feature Transform (SIFT) algorithm for feature extraction and distance calculation. The suggested method starts by extracting SIFT features from a set of photos, building a keypoints and descriptor database. By capturing the unique qualities of nearby image regions, these features enable reliable matching and retrieval. The security of private cloud data, which includes query queries, the search tree, and outsourced photographs, is another major issue. Use a feature extraction method first for integrated picture features, which are composed of fundamental components like colour and shape. In particular, because the proposed method uses a balancing index tree, it can achieve logarithmic search time. Second, the picture and query feature are encrypted using the secure inner product. Include a system for determining duplicate picture material as well. When given a query image, SIFT feature extraction is applied to it, producing keypoints and descriptors that correspond to its visual characteristics. Next, a distance computation method like Euclidean distance is used to compare the features of the database photos with the query image features. The similarity between the query image and the database images is measured through this comparison. The retrieved photos are ranked in order of similarity to the query image based on the calculated distances. Search results with lesser distances between images are deemed more similar and are displayed first. To locate visually related images in huge databases, the image retrieval system that uses SIFT feature extraction and distance calculation provides a reliable and effective solution. It contributes to developments in multimedia retrieval, visual analytics, and picture understanding by enabling applications like content-based image search, image recommendation systems, and image clustering.

Learning effective representations suitable for decision making in high-level cognitive space is crucial for visual abstract reasoning tasks. The visual system of the mammalian brain is organized into parallel networks that can be roughly classified in dichotomy as the dorsal and ventral streams. How do parallel networks learn efficient representations for cognitive tasks is still an elusive question. We propose the Information Competition Learning Network (ICNet) within a mutual information-constrained framework to learn diversified representations for visual abstract reasoning tasks. ICNet comprises a representation learning module and a rule extractor module. The representation learning module learns two complementary sets of representation under different constraints. These two sets compete to prevent from learning what the other has learned, thereby minimizing mutual predictability. Subsequently, these sets are combined synergistically and relayed to the rule extractor module, where discrete abstract rules are formed to predict the correct option. Empirical experiments consistently show that ICNet achieves superior results across several visual abstract reasoning datasets. Additionally, in Out-of-Distribution relationship reasoning benchmarks, ICNet demonstrates robust generalization ability.

Deep learning based point cloud registration: an overview

Convolutional Neural Networks have been widely applied in fault diagnosis tasks of mechanical systems due to their strong feature extraction and classification capabilities. However, they have limitations in handling global context information. Vision Transformers, by leveraging self-attention mechanisms to capture global dependencies, have shown excellent performance in many visual tasks, but often come with high computational costs. Therefore, this paper proposes a lightweight and efficient intelligent fault diagnosis method based on the fusion of Convolutional Network and Vision Transformer features (FCNVT). This method combines the local feature extraction capability of CNNs with the global dependency capturing ability of ViTs, while maintaining computational efficiency. Random overlapping sampling (ROS) techniques are used to preprocess signals, generating two-dimensional synchronized wavelet transform (SWT) images as inputs to the network. Experimental verification has shown that the proposed method achieves up to 100% classification accuracy, with the model having 7 million parameters and a computational cost of only 0.28 G, outperforming other state-of-the-art methods. Finally, a graphical user interface (GUI)-based mechanical equipment fault detection system was developed using this method, which holds positive implications for advancing the practical application of intelligent fault diagnosis in mechanical equipment.

Electrocorticography (ECoG) signals provide a valuable window into neural activity, yet their complex structure makes reliable classification challenging. This study addresses the problem by proposing a feature-selective framework that integrates multiple feature extraction techniques with statistical feature selection to improve classification performance. Power spectral density, wavelet-based features, Shannon entropy, and Hjorth parameters were extracted from ECoG signals obtained during a visual task. The most informative features were then selected using analysis of variance (ANOVA), and classification was performed with several machine learning methods, including decision trees, support vector machines, neural networks, and long short-term memory (LSTM) networks. Experimental results show that the proposed framework achieves high accuracy across individual patients as well as the combined dataset, with clear separability between classes confirmed through t-SNE visualization. In addition, analysis of selected features highlights the prominent role of electrodes located near the visual cortex, providing insights into the spatial distribution of neural activity.

Convolutional neural networks have powerful performances in many visual tasks because of their hierarchical structures and powerful feature extraction capabilities. SPD (symmetric positive definition) matrix is paid attention to in visual classification, because it has excellent ability to learn proper statistical representation and distinguish samples with different information. In this paper, a deep neural network signal detection method based on spectral convolution features is proposed. In this method, local features extracted from convolutional neural network are used to construct the SPD matrix, and a deep learning algorithm for the SPD matrix is used to detect target signals. Feature maps extracted by two kinds of convolutional neural network models are applied in this study. Based on this method, signal detection has become a binary classification problem of signals in samples. In order to prove the availability and superiority of this method, simulated and semi-physical simulated data sets are used. The results show that, under low SCR (signal-to-clutter ratio), compared with the spectral signal detection method based on the deep neural network, this method can obtain a gain of 0.5–2 dB on simulated data sets and semi-physical simulated data sets.

Large annotated datasets inevitably contain noisy labels, which poses a major challenge for training deep neural networks as they easily memorize the labels. Noise-robust loss functions have emerged as a notable strategy to counteract this issue, but it remains challenging to create a robust loss function which is not susceptible to underfitting. Through a quantitative approach, this paper explores the limited overlap between the network output at initialization and regions of non-vanishing gradients of bounded loss functions in the initial learning phase. Using these insights, we address underfitting of several noise robust losses with a novel method denoted as logit bias, which adds a real number epsilon to the logit at the position of the correct class. The logit bias enables these losses to achieve state-of-the-art results, even on datasets like WebVision, consisting of over a million images from 1000 classes. In addition, we demonstrate that our method can be used to determine optimal parameters for several loss functions – without having to train networks. Remarkably, our method determines the hyperparameters based on the number of classes, resulting in loss functions which require zero dataset or noise-dependent parameters.

Recently, emotion recognition in conversation (ERC) has become more crucial in the development of diverse Internet of Things devices, especially closely connected with users. The majority of deep learning-based methods for ERC combine the multilayer, bidirectional, recurrent feature extractor and the attention module to extract sequential features. In addition to this, the latest model utilizes speaker information and the relationship between utterances through the graph network. However, before the input is fed into the bidirectional recurrent module, detailed intrautterance features should be obtained without variation of characteristics. In this article, we propose a residual-based graph convolution network (RGCN) and a new loss function. Our RGCN contains the residual network (ResNet)-based, intrautterance feature extractor and the GCN-based, interutterance feature extractor to fully exploit the intra-inter informative features. In the intrautterance feature extractor based on ResNet, the elaborate context feature for each independent utterance can be produced. Then, the condensed feature can be obtained through an additional GCN-based, interutterance feature extractor with the neighboring associated features for a conversation. The proposed loss function reflects the edge weight to improve effectiveness. Experimental results demonstrate that the proposed method achieves superior performance compared with state-of-the-art methods.

Siamese neural network is a very powerful architecture for both feature extraction and metric learning. It usually consists of several networks that share weights. The Siamese concept is topology-agnostic and can use any neural network as its backbone. The two most popular loss functions for training these networks are the triplet and contrastive loss functions. In this paper, we propose two novel loss functions, named Fisher Discriminant Triplet (FDT) and Fisher Discriminant Contrastive (FDC). The former uses anchor-neighbor-distant triplets while the latter utilizes pairs of anchor-neighbor and anchor-distant samples. The FDT and FDC loss functions are designed based on the statistical formulation of the Fisher Discriminant Analysis (FDA), which is a linear subspace learning method. Our experiments on the MNIST and two challenging and publicly available histopathology datasets show the effectiveness of the proposed loss functions.