Sharif Digital Repository

In multi-label classification problems, each instance can simultaneously have multiple labels. Since the whole number of available labels in real-world applications tends to be (very) large, multi-label classification becomes an important challenge and recently label space dimension reduction (LSDR) methods have received attention. These methods first encode the output space to a low-dimensional latent space. Afterwards, they predict the latent space from the feature space and reconstruct the original output space using a suitable decoding method. The encoding method can be implicit which learns a code matrix in the latent space by solving an optimization problem or explicit which learns a... 

Multi-label classification with a large set of labels is a challenging task. Label-Space Dimension Reduction (LSDR) is the most popular approach that addresses this problem. LSDR methods project the high-dimensional label vectors onto a low-dimensional space that can be predicted from the feature space. Many LSDR methods assume that the training data provide complete label vector for all training samples while this assumption is usually violated particularly when label vectors are high dimensional. In this paper, we propose a probabilistic model that has an effective mechanism to handle missing and noisy labels. In the proposed Bayesian network model, a set of auxiliary random variables,... 

Traditional learning algorithms cannot perform well in scenarios where training data (source domain data) that are used to learn the model have a different distribution with test data (target domain data). The domain adaptation that intends to compensate this problem is an important capability for an intelligent agent. This paper presents a domain adaptation method which learns to adapt the data distribution of the source domain to that of the target domain where no labeled data of the target domain is available (and just unlabeled data are available for the target domain). Our method jointly learns a low dimensional representation space and an adaptive classifier. In fact, we try to find a... 

Multi-label classification has received considerable interest in recent years. Multi-label classifiers usually need to address many issues including: handling large-scale datasets with many instances and a large set of labels, compensating missing label assignments in the training set, considering correlations between labels, as well as exploiting unlabeled data to improve prediction performance. To tackle datasets with a large set of labels, embedding-based methods represent the label assignments in a low dimensional space. Many state-of-the-art embedding-based methods use a linear dimensionality reduction to map the label assignments to a low-dimensional space. However, by doing so, these... 

Multi-label classification has received considerable interest in recent years. Multi-label classifiers usually need to address many issues including: handling large-scale datasets with many instances and a large set of labels, compensating missing label assignments in the training set, considering correlations between labels, as well as exploiting unlabeled data to improve prediction performance. To tackle datasets with a large set of labels, embedding-based methods represent the label assignments in a low-dimensional space. Many state-of-the-art embedding-based methods use a linear dimensionality reduction to map the label assignments to a low-dimensional space. However, by doing so, these... 

k-medoids algorithm is a partitional, centroid-based clustering algorithm which uses pairwise distances of data points and tries to directly decompose the dataset with n points into a set of k disjoint clusters. However, k-medoids itself requires all distances between data points that are not so easy to get in many applications. In this paper, we introduce a new method which requires only a small proportion of the whole set of distances and makes an effort to estimate an upperbound for unknown distances using the inquired ones. This algorithm makes use of the triangle inequality to calculate an upper-bound estimation of the unknown distances. Our method is built upon a recursive approach to... 

Almost all of the existing domain adaptation methods assume that all test data belong to a single stationary target distribution. However, in many real world applications, data arrive sequentially and the data distribution is continuously evolving. In this paper, we tackle the problem of adaptation to a continuously evolving target domain that has been recently introduced. We assume that the available data for the source domain are labeled but the examples of the target domain can be unlabeled and arrive sequentially. Moreover, the distribution of the target domain can evolve continuously over time. We propose the Evolving Domain Adaptation (EDA) method that first finds a new feature space... 

Recently, the problem of integrating side information about classes has emerged in the learning settings like zero-shot learning. Although using multiple sources of information about the input space has been investigated in the last decade and many multi-view and multi-modal learning methods have already been introduced, the attribute learning for classes (output space) is a new problem that has been attended in the last few years. In this paper, we propose an attribute learning method that can use different sources of descriptions for classes to find new attributes that are more proper to be used as class signatures. Experimental results show that the learned attributes by the proposed... 

In many real-world applications, data contain heterogeneous input modalities (e.g., web pages include images, text, etc.). Moreover, data such as images are usually described using different views (i.e. different sets of features). Learning a distance metric or similarity measure that originates from all input modalities or views is essential for many tasks such as content-based retrieval ones. In these cases, similar and dissimilar pairs of data can be used to find a better representation of data in which similarity and dissimilarity constraints are better satisfied. In this paper, we incorporate supervision in the form of pairwise similarity and/or dissimilarity constraints into... 

One of the most studied problems in machine learning is finding reasonable constraints that guarantee the generalization of a learning algorithm. These constraints are usually expressed as some simplicity assumptions on the target. For instance, in the Vapnik–Chervonenkis (VC) theory the space of possible hypotheses is considered to have a limited VC dimension One way to formulate the simplicity assumption is via information theoretic concepts. In this paper, the constraint on the entropy H(X) of the input variable X is studied as a simplicity assumption. It is proven that the sample complexity to achieve an ∊-δ Probably Approximately Correct (PAC) hypothesis is bounded by [Formula... 

Although GAN-based methods have received many achievements in the last few years, they have not been entirely successful in generating discrete data. The most crucial challenge of these methods is the difficulty of passing the gradient from the discriminator to the generator when the generator outputs are discrete. Despite the fact that several attempts have been made to alleviate this problem, none of the existing GAN-based methods have improved the performance of text generation compared with the maximum likelihood approach in terms of both the quality and the diversity. In this paper, we proposed a new framework for generating discrete data by an adversarial approach in which there is no... 

Background: We developed transformer-based deep learning models based on natural language processing for early risk assessment of Alzheimer’s disease from the picture description test. Methods: The lack of large datasets poses the most important limitation for using complex models that do not require feature engineering. Transformer-based pre-trained deep language models have recently made a large leap in NLP research and application. These models are pre-trained on available large datasets to understand natural language texts appropriately, and are shown to subsequently perform well on classification tasks with small training sets. The overall classification model is a simple classifier on... 

Deep learning has received much attention as of the most powerful approaches for multimodal representation learning in recent years. An ideal model for multimodal data can reason about missing modalities using the available ones, and usually provides more information when multiple modalities are being considered. All the previous deep models contain separate modality-specific networks and find a shared representation on top of those networks. Therefore, they only consider high level interactions between modalities to find a joint representation for them. In this paper, we propose a multimodal deep learning framework (MDLCW) that exploits the cross weights between representation of... 

Deep generative models have achieved great success in areas such as image, speech, and natural language processing in the past few years. Thanks to the advances in graph-based deep learning, and in particular graph representation learning, deep graph generation methods have recently emerged with new applications ranging from discovering novel molecular structures to modeling social networks. This paper conducts a comprehensive survey on deep learning-based graph generation approaches and classifies them into five broad categories, namely, autoregressive, autoencoder-based, reinforcement learning-based, adversarial, and flow-based graph generators, providing the readers a detailed description... 

Deep learning-based graph generation approaches have remarkable capacities for graph data modeling, allowing them to solve a wide range of real-world problems. Making these methods able to consider different conditions during the generation procedure even increases their effectiveness by empowering them to generate new graph samples that meet the desired criteria. This paper presents a conditional deep graph generation method called SCGG that considers a particular type of structural conditions. Specifically, our proposed SCGG model takes an initial subgraph and autoregressively generates new nodes and their corresponding edges on top of the given conditioning substructure. The architecture... 

Disease prediction is a well-known classification problem in medical applications. Graph Convolutional Networks (GCNs) provide a powerful tool for analyzing the patients’ features relative to each other. This can be achieved by modeling the problem as a graph node classification task, where each node is a patient. Due to the nature of such medical datasets, class imbalance is a prevalent issue in the field of disease prediction, where the distribution of classes is skewed. When the class imbalance is present in the data, the existing graph-based classifiers tend to be biased towards the major class(es) and neglect the samples in the minor class(es). On the other hand, the correct diagnosis... 

The increased amount of multi-modal medical data has opened the opportunities to simultaneously process various modalities such as imaging and non-imaging data to gain a comprehensive insight into the disease prediction domain. Recent studies using Graph Convolutional Networks (GCNs) provide novel semi-supervised approaches for integrating heterogeneous modalities while investigating the patients’ associations for disease prediction. However, when the meta-data used for graph construction is not available at inference time (e.g., coming from a distinct population), the conventional methods exhibit poor performance. To address this issue, we propose a novel semi-supervised approach named GKD... 

Graph data structures are fundamental for studying connected entities. With an increase in the number of applications where data is represented as graphs, the problem of graph generation has recently become a hot topic. However, despite its significance, conditional graph generation that creates graphs with desired features is relatively less explored in previous studies. This paper addresses the problem of class-conditional graph generation that uses class labels as generation constraints by introducing the Class Conditioned Graph Generator (CCGG). We built CCGG by injecting the class information as an additional input into a graph generator model and including a classification loss in its... 

Computed tomography (CT) simulators are versatile tools for scanning protocol evaluation, optimization of geometrical design parameters, assessment of image reconstruction algorithms, and evaluation of the impact of future innovations attempting to improve the performance of CT scanners. Computational human phantoms (CHPs) play a key role in simulators for the radiation dosimetry and assessment of image quality tasks in the medical x-ray systems. Since the construction of patient-specific CHPs can be both difficult and time-consuming, nominal standard/reference CHPs have been established, yielding significant discrepancies in the special design and optimization demands of patient dose and...