Search

We present an SE(3) Group Convolutional Neural Network along with a series of networks with different group actions for segmentation of Diffusion Weighted Imaging data. These networks gradually incorporate group actions that are natural for this type of data, in the form of convolutions that provide equivariant transformations of the data. This knowledge provides a potentially important inductive bias and may alleviate the need for data augmentation strategies. We study the effects of these actions on the performances of the networks by training and validating them using the diffusion data from the Human Connectome project. Unlike previous works that use Fourier-based convolutions, we implement direct convolutions, which are more lightweight. We show how incorporating more actions - using the SE(3) group actions - generally improves the performances of our segmentation while limiting the number of parameters that must be learned.

PurposeApplying machine learning techniques to magnetic resonance diffusion-weighted imaging (DWI) data is challenging due to the size of individual data samples and the lack of labeled data. It is possible, though, to learn general patterns from a very limited amount of training data if we take advantage of the geometry of the DWI data. Therefore, we present a tissue classifier based on a Riemannian deep learning framework for single-shell DWI data.ApproachThe framework consists of three layers: a lifting layer that locally represents and convolves data on tangent spaces to produce a family of functions defined on the rotation groups of the tangent spaces, i.e., a (not necessarily continuous) function on a bundle of rotational functions on the manifold; a group convolution layer that convolves this function with rotation kernels to produce a family of local functions over each of the rotation groups; a projection layer using maximization to collapse this local data to form manifold based functions.ResultsExperiments show that our method achieves the performance of the same level as state-of-the-art while using way fewer parameters in the model (ConclusionsThis work extended convolutional neural networks to Riemannian manifolds, and it shows the potential in understanding structural patterns in the brain, as well as in aiding manual data annotation. Purpose: Applying machine learning techniques to magnetic resonance diffusion-weighted imaging (DWI) data is challenging due to the size of individual data samples and the lack of labeled data. It is possible, though, to learn general patterns from a very limited amount of training data if we take advantage of the geometry of the DWI data. Therefore, we present a tissue classifier based on a Riemannian deep learning framework for single-shell DWI data.Approach: The framework consists of three layers: a lifting layer that locally represents and convolves data on tangent spaces to produce a family of functions defined on the rotation groups of the tangent spaces, i.e., a (not necessarily continuous) function on a bundle of rotational functions on the manifold; a group convolution layer that convolves this function with rotation kernels to produce a family of local functions over each of the rotation groups; a projection layer using maximization to collapse this local data to form manifold based functions.Results: Experiments show that our method achieves the performance of the same level as state-of-the-art while using way fewer parameters in the model (< 10 % ). Meanwhile, we conducted a model sensitivity analysis for our method. We ran experiments using a proportion (69.2%, 53.3%, and 29.4%) of the original training set and analyzed how much data the model needs for the task. Results show that this does reduce the overall classification accuracy mildly, but it also boosts the accuracy for minority classes.Conclusions: This work extended convolutional neural networks to Riemannian manifolds, and it shows the potential in understanding structural patterns in the brain, as well as in aiding manual data annotation.

In order to solve the problem of the lack of information in the evaluation of university students' learning behavior with single data, a multi-source data fusion evaluation of university students' learning behavior is proposed to explore the application value of data fusion technology in teaching. First use Spearman to calculate the correlation between different data sources, calculate the relationship weight and optimization weight of different data sources, and then comprehensively weight for fusion calculation. The fusion calculation of students' comprehensive learning behavior evaluation through comprehensive weighted fusion can not only balance the differences in learning behavior evaluation of students under different teaching methods, but also integrate the advantages and disadvantages of different learning methods. Experiments verify that the comprehensive weighted data fusion algorithm is effective, which can provide help for a more comprehensive analysis of the evaluation of college students' learning behavior and provide a basis for adapting to their own learning.

Additional file 2. The percentage of cells in each quadrant in all apoptosis experiments.

Additional file 3. Raw data from three repeated experiments.

Acridone derivatives with electron-rich triphenylamine functionalized at the amino position were synthesized and their properties were experimentally and computationally investigated. The single crystal structure analysis revealed that the π–π interaction of acridone and the formation of hydrogen bonds of carbonyl and the hydrogen atoms of the pending phenyl ring were crucial in the determination of molecular packing in the crystalline state. An intramolecular charge transfer (ICT) process was observed between acridone and triphenylamine even with reduced conjugation by the nitrogen atom of acridone. Tuneable aggregation induced emissions with blue and green fluorescence were found due to the different aggregation state and particle size, which varied according to the water content in THF. Furthermore, the size of the spacer between acridone and the appended amine was also important in adjusting the property of aggregation induced emission or aggregation caused quenching in the solid state.

The era of big data brings together a wealth of rich information resources, bringing unprecedented innovation power, opportunities and possibilities to all walks of life, and promoting the development and evolution of the era. In the current development process of college student associations, to implement scientific and effective talent management, big data technology must be introduced to assist the management of student associations. Make full use of big data technology as the most basic strategic resource and effective means to transform the management of student associations from static and result-based management to dynamic and process-oriented. This article starts from the concept of big data, combines the definition and characteristics of college student associations, uses survey research methods, literature research methods, and comparative analysis methods, and issues a questionnaire on "Student Association Management and Association Activity Survey". A special investigation, analysis and sorting of the activities carried out were carried out. Among them, the efficiency of resource sharing in student club management is relatively low, accounting for 52.35%. The application strategies of big data are studied from the aspects of student club management system, club evaluation mechanism, student activity needs, community resource sharing and intercommunication.

Hybrid teaching is widely used in colleges and universities in the network teaching platform, can effectively make up for the shortcomings of traditional classroom teaching in colleges and universities at present, but because of the network teaching lack of the guidance of system theory and teaching practice, teaching in colleges and universities network platform will inevitably be in the application of some technical problems, so you need to take the design of the new teaching system. The purpose of this paper is to design a mixed online and offline teaching system based on network information technology, and to provide technical Suggestions for the optimization of mixed teaching methods. This paper designs a hybrid teaching system based on information technology and conducts an empirical study through teaching experiment. The experimental subjects were sophomores in a certain university, and two classes were randomly selected for the teaching experiment. The results showed that the students in the experimental group generally had a high evaluation on the teaching platform, and 93.1% of the students thought the teaching system was very useful. From the results, the design of the new system can help improve the online and offline hybrid teaching system.

Using multiple features in appearance modeling has shown to be effective for visual tracking. In this paper, we dynamically measured the importance of different features and proposed a robust tracker with the weighted features. By doing this, the dictionaries are improved in both reconstructive and discriminative way. We extracted multiple features of the target, and obtained multiple sparse representations, which plays an essential role in the classification issue. After learning independent dictionaries for each feature, we then implement weights to each feature dynamically, with which we select the best candidate by a weighted joint decision measure. Experiments have shown that our method outperforms several recently proposed trackers.

Fast Fourier transform (FFT) and inverse FFT (IFFT) are used in orthogonal frequency-division multiplexing (OFDM) systems for efficiently converting frequency domain signals that carry discrete information symbols from or to time domain sequences, respectively. This discreteness of the frequency-domain symbol set in context of FFT is of great interest in this paper. In such a paradigm, unfortunately, sparse data corruption in the time domain sequence leads to catastrophic signal degradation in FFT due to its natural noise spread effect, and little can be mitigated by using error control coding schemes (ECC). Potential causes of sparse errors include timedomain sample loss, impulsive noise, and circuit failure in integrated circuits. In this paper, we first characterize the impact of sparse severe errors in the context of FFT, as opposed to that of commonly considered additive white Gaussian noise (AWGN). Based on observation of distinct impact, a series of computationally efficient algorithms are proposed for detecting and mitigating such errors in lieu of ECC. Simulation results show that our proposed algorithms can effectively combat various sparse errors and dramatically improve the performance.