Search

Purpose: To train deep learning models to differentiate benign and malignant breast tumors in ultrasound images, we need to collect many training samples with clear labels. In general, biopsy results can be used as benign/malignant labels. However, most clinical samples generally do not have biopsy results. Previous works have proposed generating benign/malignant labels according to Breast Imaging, Reporting and Data System (BI-RADS) ratings. However, this approach will cause noisy labels, which means that the benign/malignant labels produced from BI-RADS diagnoses may be inconsistent with the ground truths. Consequently, deep models will overfit the noisy labels and hence obtain poor generalization performance. In this work, we mainly focus on how to reduce the negative effect of noisy labels when they are used to train breast tumor classification models., Methods: We propose an effective approach called noise filter network (NF-Net) to address the problem of noisy labels when training breast tumor classification models. Specifically, to prevent deep models from overfitting the noisy labels, we propose incorporating two softmax layers for classification. Additionally, to strengthen the effect of clean labels, we design a teacher-student module for distilling the knowledge of clean labels., Results: We conduct extensive comparisons with the existing works on addressing noisy labels. Our method achieves a classification accuracy of 73%, with a precision of 69%, recall of 80%, and F1-score of 0.74. This result is significantly better than those of the existing state-of-the-art works on addressing noisy labels., Conclusions: This work provides a means to overcome the label shortage problem in training breast tumor classification models. Specifically, we can generate benign/malignant labels according to the BI-RADS ratings. Although this approach will cause noisy labels, the design of NF-Net can effectively reduce the negative effect of such labels., (© 2019 American Association of Physicists in Medicine.)

Recommendation efficiency and data sparsity problems have been regarded as two challenges of improving performance for online recommendation. Most of the previous related work focus on improving recommendation accuracy instead of efficiency. In this paper, we propose a Deep Pairwise Hashing (DPH) to map users and items to binary vectors in Hamming space, where a user's preference for an item can be efficiently calculated by Hamming distance, which significantly improves the efficiency of online recommendation. To alleviate data sparsity and cold-start problems, the user-item interactive information and item content information are unified to learn effective representations of items and users. Specifically, we first pre-train robust item representation from item content data by a Denoising Auto-encoder instead of other deterministic deep learning frameworks; then we finetune the entire framework by adding a pairwise loss objective with discrete constraints; moreover, DPH aims to minimize a pairwise ranking loss that is consistent with the ultimate goal of recommendation. Finally, we adopt the alternating optimization method to optimize the proposed model with discrete constraints. Extensive experiments on three different datasets show that DPH can significantly advance the state-of-the-art frameworks regarding data sparsity and item cold-start recommendation. [ABSTRACT FROM AUTHOR]

In recent years, malware with strong concealment uses encrypted protocol to evade detection. Thus, encrypted traffic identification can help security analysts to be more effective in narrowing down those encrypted network traffic. Existing methods are protocol independent, such as statistical-based and machine-learning-based approaches. Statistical-based approaches, however, are confined to payload length and machine-learning-based approaches have a low recognition rate for encrypted traffic using undisclosed protocols. In this paper, we proposed a heuristic statistical testing (HST) approach that combines both statistics and machine learning and has been proved to alleviate their respective deficiencies. We manually selected four randomness tests to extract small payload features for machine learning to improve real-time performances. We also proposed a simple handshake skipping method called HST-R to increase the classification accuracy. We compared our approach with other identification approaches on a testing dataset consisting of traffic that uses two known, two undisclosed, and one custom cryptographic protocols. Experimental results showed that HST-R performs better than other traditional coding-based, entropy-based, and ML-based approaches. We also showed that our handshake skipping method could generalize better for unknown cryptographic protocols. Finally, we also conducted experimental comparisons among different classification algorithms. The results showed that C4.5, with our method, has the highest identification accuracy for secure sockets layer and secure shell traffic. [ABSTRACT FROM AUTHOR]

In many machine learning applications such as computer-aided diagnosis, gene sequence analysis or natural language processing, categorical data appears. For small-scale data set with high dimensions, since relatively small proportion of possible categorical configurations are covered by training samples, conventional methods based on frequency information such as Dirichlet Compound Multinomial distribution usually runs into problems of over-fitting. Latent gaussian process is an effective bayesian non-parametric technique for categorical data modeling, which was proposed as an unsupervised method to embed unlabelled categorical data into a continuous and low-dimensional space through gaussian process. As a probabilistic generative model, latent gaussian process owns the ability of density estimation. In this paper, we propose a generative classification model as a supervised method for labelled categorical data, in which we use latent gaussian process to estimate the class-conditional densities. Since the complexity of gaussian process model can adapt to the size of training data, our method is able to effectively model small-sale categorical data. Experimental results show that our proposal can achieve better classification performance compared with other classification models for categorical data. [ABSTRACT FROM AUTHOR]

Traffic sign recognition is a classification problem that poses challenges for computer vision and machine learning algorithms. Although both computer vision and machine learning techniques have constantly been improved to solve this problem, the sudden rise in the number of unlabeled traffic signs has become even more challenging. Large data collation and labeling are tedious and expensive tasks that demand much time, expert knowledge, and fiscal resources to satisfy the hunger of deep neural networks. Aside from that, the problem of having unbalanced data also poses a greater challenge to computer vision and machine learning algorithms to achieve better performance. These problems raise the need to develop algorithms that can fully exploit a large amount of unlabeled data, use a small amount of labeled samples, and be robust to data imbalance to build an efficient and high-quality classifier. In this work, we propose a novel semi-supervised classification technique that is robust to small and unbalanced data. The framework integrates weakly-supervised learning and self-training with self-paced learning to generate attention maps to augment the training set and utilizes a novel pseudo-label generation and selection algorithm to generate and select pseudo-labeled samples. The method improves the performance by: (1) normalizing the class-wise confidence levels to prevent the model from ignoring hard-to-learn samples, thereby solving the imbalanced data problem; (2) jointly learning a model and optimizing pseudo-labels generated on unlabeled data; and (3) enlarging the training set to satisfy the hunger of deep learning models. Extensive evaluations on two public traffic sign recognition datasets demonstrate the effectiveness of the proposed technique and provide a potential solution for practical applications. [ABSTRACT FROM AUTHOR]

Generative adversarial network (GAN) has shown profound power in machine learning. It inspires many researchers from other fields to create powerful tools for various tasks, including quantum state preparation, quantum circuit translation, and so on. It is known as classical techniques cannot efficiently simulate the quantum system, and the existing works haven't investigated the quantum version of maximum mean discrepancy as the metric in learning models and applied it to quantum data. In this paper, we propose a metric named quantum maximum mean discrepancy (qMMD), which can be used to measure the distance between quantum data in Hilbert space. Based on the qMMD, we then design a quantum generative adversarial model, named qMMD-GAN, under the hybrid quantum–classical methods. We also provide the construction of qMMD-GAN that can be easily implemented on a quantum device. We demonstrate the power of our qMMD-GAN by applying it to a crucial real-world application that is generating an unknown quantum state. Our numerical experiments show that qMMD-GAN has a competitive performance compared to existing results. We believe that the hybrid-based models will not only be applied to physics research but provide a new direction for improving classical data processing tasks. [ABSTRACT FROM AUTHOR]