Traditional collaborative filtering algorithm relies on score matrix to generate prediction, fails to mine potential features related to item genres, which leads to a low recommendation accuracy. A denoising autoencod...
详细信息
Traditional collaborative filtering algorithm relies on score matrix to generate prediction, fails to mine potential features related to item genres, which leads to a low recommendation accuracy. A denoising autoencoder based collaborative filtering recommendation algorithm combining item types is proposed. Firstly, combining item score data and genres data, the potential features of the items are extracted by denoising autoencoder. Then the prediction is generated by the collaborative filtering algorithm. Experiments on MovieLens dataset show that the new algorithm can mine the potential features of item genres comprehensively, and improve the recommendation accuracy.
Knowledge graph completion (KGC) involves enhancing existing factual knowledge by automatically inferring missing links between entities. However, they are limited to inferring relations for entity-pairs that have cor...
详细信息
Knowledge graph completion (KGC) involves enhancing existing factual knowledge by automatically inferring missing links between entities. However, they are limited to inferring relations for entity-pairs that have corresponding samples in the graph. In this paper, we endeavor to identify relations without training examples. To this end, we learn relation embeddings from textual descriptions and utilize a conditional variational autoencoder (C-VAE) to connect these descriptions with the corresponding entity-pair embeddings. However, two problems still persist: first, the embeddings of different relations are entangled, leading to the inability to discriminate between different relations;second, entity-pair embeddings are contaminated with impurities that decrease the accuracy of predictions for novel relations. This study introduces contrastive learning for zero-shot relational learning (CZRL). To better distinguish between different relations, we train a feature encoder with specifically designed contrastive losses. To eliminate noise, we propose a contrastive denoising autoencoder module to isolate the relevant information of entity-pair embeddings from irrelevant information. On two public datasets - Wiki, and NELL - the proposed model demonstrates the performance improved at least 10% compared to baseline models based on evaluation metrics such as MRR, Hit@1, and Hit@5.
The detection of wafer faults in early process steps through monitoring and analyzing multivariate process trace data contribute to wafer yield improvements. Standard classification algorithms have been generally used...
详细信息
The detection of wafer faults in early process steps through monitoring and analyzing multivariate process trace data contribute to wafer yield improvements. Standard classification algorithms have been generally used for fault detection and classification (FDC). However, this approach can cause information loss while extracting statistical features from the trace data and cannot consider class imbalance situations where much fewer faulty wafers are generated than normal wafers. In addition, the approach does not consider normal wafer-to-wafer (W2W) variations and sensor noise inherent in the trace data. These drawbacks significantly degrade FDC performance. This paper proposes a method that builds an FDC model only with trace data of normal wafers in which W2W variations and sensor noise exist. The one-class FDC method detects the occurrence of abnormal trace patterns that cause wafer faults by removing W2W variations and sensor noise from raw traces by using denoising autoencoders, and this method finds the fault-introducing process parameters with the occurrence times. In experiments using the trace data of etch and chemical vapor deposition processes, the proposed method exhibited 1% and 6% higher performance than the best-performing method among comparison methods in terms of the geometric mean of the normal and fault detection accuracies.
Datasets exist in real life in many formats (audio, music, image,...). In our case, we have them from various sources mixed together. Our mixtures represent noisy audio data that need to be extracted (features), compr...
详细信息
Datasets exist in real life in many formats (audio, music, image,...). In our case, we have them from various sources mixed together. Our mixtures represent noisy audio data that need to be extracted (features), compressed and analysed in order to be presented in a standard way. The resulted data will be used for the Blind Source Separation task. In this paper, we deal with two types of autoencoders: convolutional and denoising. The novelty of our work is to reconstruct the audio signal in the output of the neural network after extracting the meaningful features that present the pure and the powerful information. Simulation results show a great performance, yielding of 87% for the reconstructed signals that will be included in the automated system used for real word applications.
This paper addresses important challenges in wind energy prediction caused by outliers in wind data, which distort the wind turbine power curve and lead to inaccurate performance assessments and suboptimal operation s...
详细信息
This paper addresses important challenges in wind energy prediction caused by outliers in wind data, which distort the wind turbine power curve and lead to inaccurate performance assessments and suboptimal operation strategies. The major difficulty here is detecting and eliminating these outliers from complex wind datasets, as inaccurate data can significantly impact forecasting and related activities. To overcome this challenge, the paper proposes a hybrid model combining fuzzy C-means clustering, Mahalanobis distance, and Artificial Neural Networks (ANN) to detect and remove outliers far more accurately than any individual method or other traditional hybrid method, decreasing false alarms and misses. It improves data quality and boosts the reliability of turbine performance analysis, resource assessment, and forecasting, supporting more efficient and sustainable wind-power operations. The results show (1) that the proposed hybrid model achieves 15.4% more accuracy than the other traditional hybrid models in detecting and removing outliers. (2) The proposed hybrid model gives an overall ≈ 116.1% improvement in outlier-detection accuracy over the individual models. (3) Adding the ANN to the proposed hybrid model boosts the outlier-detection accuracy to about a 69.5% relative improvement. (4) Detecting and cleaning outliers by the proposed hybrid model cuts the RMSE from 2.38 to 1.27, reducing prediction error by 46.6%. (5) The advanced hybrid model used in this study for comparison purposes achieves nearly identical accuracy to the proposed hybrid model; it reduces RMSE by ∼0.015 and MAPE by ∼0.04 pp and boosts R² by ∼0.001 while maintaining almost perfect outlier detection (99 % vs. 100 %). Although the advanced model offers a marginal edge in reconstruction quality, the lightweight, scalable proposed hybrid model remains better appropriate for real-world deployment due to its lower computational overhead and more straightforward maintenance.
In order to suppress the late reverberation in the spectral domain, many single-channel dereverberation techniques rely on an estimate of the late reverberation power spectral density (PSD). In this paper, we propose ...
详细信息
ISBN:
(纸本)9781510872219
In order to suppress the late reverberation in the spectral domain, many single-channel dereverberation techniques rely on an estimate of the late reverberation power spectral density (PSD). In this paper, we propose a novel approach to late reverberation PSD estimation using a denoising autoencoder (DA), which is trained to learn a mapping from the microphone signal PSD to the late reverberation PSD. Simulation results show that the proposed approach yields a high PSD estimation accuracy and generalizes well to unseen data. Furthermore, simulation results show that the proposed DA-based PSD estimate yields a higher PSD estimation accuracy and a similar dereverberation performance than a state-of-the-art statistical PSD estimate, which additionally also requires knowledge of the reverberation time.
In the industrial applications like fault diagnosis and health management,monitoring data generally reaches sequentially in a streaming *** recognize fault occurrence in real time without system halt,it is necessary t...
详细信息
In the industrial applications like fault diagnosis and health management,monitoring data generally reaches sequentially in a streaming *** recognize fault occurrence in real time without system halt,it is necessary to improve the accuracy and stability of anomaly detection with streaming *** solve this problem,a new online anomaly detection method with streaming data is proposed based on fine-grained feature ***,to conduct fine-grained decomposition of features,a denoising autoencoder network is run to extract multiple-dimensional deep features of online data in the initial period of normal ***,a forecasting model with tensor Tucker decomposition and ARIMA is conducted to predict the fluctuation trend of all feature ***,the deviation degree between the prediction values and sequentially-arrived data is calculated,and an alarm threshold is built according to the 95%confidence interval of the maximum *** the anomalous state data can be detected in real *** paper adopts the problem of bearing early fault online detection as an example,and run comparative experiments on the IEEE PHM Challenge 2012 bearing *** results show that the proposed method has good detection accuracy and is with no false alarm,while the model training does not rely on any offline *** the proposed method is applicable to the problem of online anomaly detection
In most recommender systems, the data of user feedbacks are usually represented with a set of discrete values, which are difficult to exactly describe users' interests. This problem makes it not easy to exactly mo...
详细信息
In most recommender systems, the data of user feedbacks are usually represented with a set of discrete values, which are difficult to exactly describe users' interests. This problem makes it not easy to exactly model users' latent preferences for recommendation. Intuitively, a basic idea for this issue is to predict continuous values through a trained model to reveal users' essential feedbacks, and then make use of the generated data to retrain another model to learn users' preferences. However, since these continuous data are generated by an imperfect model which are trained by discrete data, there exists a lot of noise among the generated data. This problem may have a severe adverse impact on the performance. Towards this problem, we propose a novel Enhanced Collaborative autoencoder (ECAE) to learn robust information from generated soft data with the technique of knowledge distillation. First, we propose a tightly coupled structure to incorporate the generation and retraining stages into a unified framework. So that the generated data can be fine tuned to reduce the noise by propagating training errors of retraining network. Second, for that each unit of the generated data contains different level of noise, we propose a novel distillation layer to balance the influence of noise and knowledge. Finally, we propose to take both predict results of generation and retraining network into account to make final recommendations for each user. The experimental results on four public datasets for top-N recommendation show that the ECAE model performs better than several state-of-the-art algorithms on metrics of MAP and NDCG. (C) 2019 Elsevier B.V. All rights reserved.
Automatic modulation classification plays an important role in many fields to identify the modulation type of wireless signals in order to recover signals by demodulation. In this paper, we contribute to explore the s...
详细信息
Automatic modulation classification plays an important role in many fields to identify the modulation type of wireless signals in order to recover signals by demodulation. In this paper, we contribute to explore the suitable architecture of deep learning method in the domain of communication signal recognition. Based on architecture analysis of the convolutional neural network, we used real signal data generated by instrument as dataset, and achieved compatible recognition accuracy of modulation classification compared with several representative structure. We state that the deeper network architecture is not suitable for the signal recognition due to its different characteristic. In addition, we also discuss the difficult of training algorithm in deep learning methods and employ the transfer learning method in order to reap the benefits, which stabilize the training process and lift the performance. Finally, we adopt the denoising autoencoder to preprocess the received data and provide the ability to resist finite perturbations of the input. It contributes to a higher recognition accuracy and it also provide a new idea to design the denoising modulation recognition model.
作者：
Sonoda, ShoMurata, NoboruRIKEN
Ctr Adv Intelligence Project Chuo Ku 1-4-1 Nihonbashi Tokyo 1030027 Japan Waseda Univ
Sch Adv Sci & Engn Shinjuku Ku 3-4-1 Okubo Tokyo 1698555 Japan
We investigated the feature map inside deep neural networks (DNNs) by tracking the transport map. We are interested in the role of depth - why do DNNs perform better than shallow models? - and the interpretation of DN...
详细信息
We investigated the feature map inside deep neural networks (DNNs) by tracking the transport map. We are interested in the role of depth - why do DNNs perform better than shallow models? - and the interpretation of DNNs- what do intermediate layers do? Despite the rapid development in their application, DNNs remain analytically unexplained because the hidden layers are nested and the parameters are not faithful. Inspired by the integral representation of shallow NNs, which is the continuum limit of the width, or the hidden unit number, we developed the flow representation and transport analysis of DNNs. The flow representation is the continuum limit of the depth, or the hidden layer number, and it is specified by an ordinary differential equation ( ODE) with a vector field. We interpret an ordinary DNN as a transport map or an Euler broken line approximation of the flow. Technically speaking, a dynamical system is a natural model for the nested feature maps. In addition, it opens a new way to the coordinate-free treatment of DNNs by avoiding the redundant parametrization of DNNs. Following Wasserstein geometry, we analyze a flow in three aspects: dynamical system, continuity equation, and Wasserstein gradient flow. A key finding is that we specified a series of transport maps of the denoising autoencoder (DAE), which is a cornerstone for the development of deep learning. Starting from the shallow DAE, this paper develops three topics: the transport map of the deep DAE, the equivalence between the stacked DAE and the composition of DAEs, and the development of the double continuum limit or the integral representation of the flow representation. As partial answers to the research questions, we found that deeper DAEs converge faster and the extracted features are better;in addition, a deep Gaussian DAE transports mass to decrease the Shannon entropy of the data distribution. We expect that further investigations on these questions lead to the development of an interpre
暂无评论