In this paper, we present a deep learning based modulation scheme for chaotic orthogonal frequency division multiplex (OFDM) transmissions over non-contiguous frequency bands of cognitive radio systems. In cognitive r...
详细信息
ISBN:
(纸本)9781538661192
In this paper, we present a deep learning based modulation scheme for chaotic orthogonal frequency division multiplex (OFDM) transmissions over non-contiguous frequency bands of cognitive radio systems. In cognitive radio systems, the users access the spectrum bands dynamically and the corresponding channel characteristics also changes. Different from the traditional modulation scheme that uses the fixed mapping pattern to modulate the signals, we propose to apply the deep learning method to build up the constellations intelligently. Based on the autoencoder architecture of deep learning, we construct the constellation mapping and demapping patterns adaptively with the aim to minimize the bit error rate (BER) over the dynamically changing non-contiguous channels. Simulation results over additive white Gaussian noise (AWGN) channel and Rayleigh fading channel show that our proposed system achieves better BER performances for legitimate receivers when compared with the conventional modulation schemes. In addition, the presented scheme remains the high security performance thanks to the usage of the chaotic sequences.
Crop classification maps based on high resolution remote sensing data are essential for supporting sustainable land management. The most challenging problems for their producing are collecting of ground based training...
详细信息
ISBN:
(纸本)9781538671504
Crop classification maps based on high resolution remote sensing data are essential for supporting sustainable land management. The most challenging problems for their producing are collecting of ground based training and validation datasets, non-regular satellite data acquisition and cloudiness. To increase the efficiency of ground data utilization it is important to develop classifiers able to be trained on the data collected in the previous year. In this study, we propose new deep learning method for providing crop classification maps using in-situ data that has been collected in the previous year. Main idea of the study is to utilize deep learning approach based on sparse autoencoder. At the first stage it is trained on satellite data only and then neural network fine-tuning is conducted based on in-situ data form the previous year. Taking into account that collecting ground truth data is very time consuming and challenging task, the proposed approach allows us to avoid necessity for annual collecting in-situ data for the same territory. Experimental results for the territory of Ukraine show that this technique is rather efficient and provides reliable crop classification maps with overall accuracy higher than 85.9%.
We consider an online decision making setting known as contextual bandit problem, and propose an approach for improving contextual bandit performance by using an adaptive feature extraction (representation learning) b...
详细信息
ISBN:
(纸本)9781538692882
We consider an online decision making setting known as contextual bandit problem, and propose an approach for improving contextual bandit performance by using an adaptive feature extraction (representation learning) based on online clustering. Our approach starts with an off-line pre-training on unlabeled history of contexts (which can be exploited by our approach, but not by the standard contextual bandit), followed by an online selection and adaptation of encoders. Specifically, given an input sample (context), the proposed approach selects the most appropriate encoding function to extract a feature vector which becomes an input for a contextual bandit, and updates both the bandit and the encoding function based on the context and on the feedback (reward). Our experiments on a variety of datasets, and both in stationary and non-stationary environments of several kinds demonstrate clear advantages of the proposed adaptive representation learning over the standard contextual bandit based on "raw" input contexts.
We present a feature engineering pipeline for the construction of musical signal characteristics, to be used for the design of a supervised model for musical genre identification. The key idea is to extend the traditi...
详细信息
ISBN:
(纸本)9781538646588
We present a feature engineering pipeline for the construction of musical signal characteristics, to be used for the design of a supervised model for musical genre identification. The key idea is to extend the traditional two-step process of extraction and classification with additive stand-alone phases which are no longer organized in a waterfall scheme. The whole system is realized by traversing backtrack arrows and cycles between various stages. In order to give a compact and effective representation of the features, the standard early temporal integration is combined with other selection and extraction phases: on the one hand, the selection of the most meaningful characteristics based on information gain, and on the other hand, the inclusion of the nonlinear correlation between this subset of features, determined by an autoencoder. The results of the experiments conducted on GTZAN dataset reveal a noticeable contribution of this methodology towards the model's performance in classification task.
Automatic document summarization is used to extract or generate short, but rich snippets which represent well the essential meaning and key information contained in documents. Classical approaches of extractive summar...
详细信息
ISBN:
(纸本)9781538670941
Automatic document summarization is used to extract or generate short, but rich snippets which represent well the essential meaning and key information contained in documents. Classical approaches of extractive summarization mostly rely on bag-of-words models or a graph representation reflecting word neighbourhood information to obtain a ranking of the sentences of the document. The higher the rank, the more relevant is the sentence for the summary. Given word-embeddings have recently been shown to represent semantic meaning of individual words in natural languages, intuitively it seems to be a straightforward way to carry out such a ranking in the latent space reflecting the meaning (semantics), not just the form (syntax). Vector operations in the semantic space have been shown to be highly consistent with alternations of the meaning reflected by such operations (i.e. analogical reasoning tasks). In the present paper we show that simply cumulatively adding semantic space vectors to represent sentence level meaning already yields comparable performance to the state-of-the-art in document summarization. The additive property of semantic representations for high number of component words is considered an important and promising outcome of this research for cognitive infocommunication applications.
Public health surveillance by traditional means is a costly and time consuming process. Today, the widespread use of social media has enabled researchers to study different aspects of life such as health, lifestyle, e...
详细信息
ISBN:
(纸本)9781538676417
Public health surveillance by traditional means is a costly and time consuming process. Today, the widespread use of social media has enabled researchers to study different aspects of life such as health, lifestyle, etc. Anonymous postings on these forums enable people to benefit from the collective experience of others facing similar problems. To effectively discern target data from the outliers in a web corpus, an efficient mechanism is required. Traditional approaches such as keyword-based filtering results in the loss of relevant data due to limited vocabulary and lack of contextual information. In this paper, we present a data filtration framework based on Long short-term memory (LSTM) recurrent neural network model for one-class text classification. We compare similarity of regenerated texts using this model for each disease with the original text using Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metric for outlier filtration and classification. Optimal value of ROUGE similarity threshold is determined by introducing an optimization parameter that minimizes the misclassification rate. Leveraging data from three major online health forums, we show that our classification technique outperforms keyword-based filtering and conventional approach of multi-class text classification. Our classification technique can be effectively used for online social networks, search engines, and online recommender systems.
This paper describes the technology behind a performance where human dancers interact with an "artificial" performer projected on a screen. The system learns movement patterns from the human dancers in real ...
详细信息
ISBN:
(数字)9783319775838
ISBN:
(纸本)9783319775838;9783319775821
This paper describes the technology behind a performance where human dancers interact with an "artificial" performer projected on a screen. The system learns movement patterns from the human dancers in real time. It can also generate novel movement sequences that go beyond what it has been taught, thereby serving as a source of inspiration for the human dancers, challenging their habits and normal boundaries and enabling a mutual exchange of movement ideas. It is central to the performance concept that the system's learning process is perceivable for the audience. To this end, an autoencoder neural network is trained in real time with motion data captured live on stage. As training proceeds, a "pose map" emerges that the system explores in a kind of improvisational state. The paper shows how this method is applied in the performance, and shares observations and lessons made in the process.
Electroencephalography (EEG)-based emotion recognition attempts to detect the affective states of humans directly via spontaneous EEG signals, bypassing the peripheral nervous system. In this thesis, we explore variou...
详细信息
Electroencephalography (EEG)-based emotion recognition attempts to detect the affective states of humans directly via spontaneous EEG signals, bypassing the peripheral nervous system. In this thesis, we explore various machine learning techniques for EEG-based emotion recognition, and focus on the three research gaps outlined as follows. 1. Stable feature selection for recalibration-less affective Brain-Computer Interfaces 2. Cross-subject transfer learning for calibration-less affective Brain- Computer Interfaces 3. Unsupervised feature learning for affective Brain-Computer Interfaces We propose several novel methods in this thesis to address the three research gaps and validate our proposed methods by experiments. Extensive comparisons between our methods and other existing methods justify the advantages of our methods.
One approach in training a deep neural network to perform effectively is to do unsupervised pretraining on each layer, followed by fine-tuning the whole network. A common way is to train an unsupervised model of neura...
详细信息
ISBN:
(纸本)9781538626337
One approach in training a deep neural network to perform effectively is to do unsupervised pretraining on each layer, followed by fine-tuning the whole network. A common way is to train an unsupervised model of neural network such as restricted Boltzmann machines or autoencoders and stack them on top of another. Although these unsupervised pretraining approaches yield good performance, relying on back-propagation, due to iterative learning process, they still suffer from a long pretraining time. Extreme learning machine (ELM) is an analytical training approach which is extremely fast and gives a solution with a good generalization performance. In this paper, we apply a new ELM based unsupervised learning, named backward ELM based autoencoder (BELM-AE), to pretrain each layer of a neural network before using a back-propagation based learning algorithm to fine-tune the whole network. Experimental results show that the new pretraining method requires significantly shorter training time and also yields better testing performance on various datasets.
Increasing demand for high field magnetic resonance (MR) scanner indicates the need for high-quality MR images for accurate medical diagnosis. However, cost constraints, instead, motivate a need for algorithms to enha...
详细信息
ISBN:
(纸本)9783030008895;9783030008888
Increasing demand for high field magnetic resonance (MR) scanner indicates the need for high-quality MR images for accurate medical diagnosis. However, cost constraints, instead, motivate a need for algorithms to enhance images from low field scanners. We propose an approach to process the given low field (3T) MR image slices to reconstruct the corresponding high field (7T-like) slices. Our framework involves a novel architecture of a merged convolutional autoencoder with a single encoder and multiple decoders. Specifically, we employ three decoders with random initializations, and the proposed training approach involves selection of a particular decoder in each weight-update iteration for back propagation. We demonstrate that the proposed algorithm outperforms some related contemporary methods in terms of performance and reconstruction time.
暂无评论