Dynamical Recurrent Networks for Sequential Data Processing

Author(s):  
Stefan C. Kremer ◽  
John F. Kolen
2022 ◽  
pp. 1-13
Author(s):  
Denis Paperno

Abstract Can recurrent neural nets, inspired by human sequential data processing, learn to understand language? We construct simplified datasets reflecting core properties of natural language as modeled in formal syntax and semantics: recursive syntactic structure and compositionality. We find LSTM and GRU networks to generalise to compositional interpretation well, but only in the most favorable learning settings, with a well-paced curriculum, extensive training data, and left-to-right (but not right-to-left) composition.


Author(s):  
Cheng Yang ◽  
Jian Tang ◽  
Maosong Sun ◽  
Ganqu Cui ◽  
Zhiyuan Liu

Information diffusion prediction is an important task which studies how information items spread among users. With the success of deep learning techniques, recurrent neural networks (RNNs) have shown their powerful capability in modeling information diffusion as sequential data. However, previous works focused on either microscopic diffusion prediction which aims at guessing the next influenced user or macroscopic diffusion prediction which estimates the total numbers of influenced users during the diffusion process. To the best of our knowledge, no previous works have suggested a unified model for both microscopic and macroscopic scales. In this paper, we propose a novel multi-scale diffusion prediction model based on reinforcement learning (RL). RL incorporates the macroscopic diffusion size information into the RNN-based microscopic diffusion model by addressing the non-differentiable problem. We also employ an effective structural context extraction strategy to utilize the underlying social graph information. Experimental results show that our proposed model outperforms state-of-the-art baseline models on both microscopic and macroscopic diffusion predictions on three real-world datasets.


Author(s):  
Jia Wang ◽  
Tong Sun ◽  
Benyuan Liu ◽  
Yu Cao ◽  
Hongwei Zhu

Financial markets are a complex dynamical system. The complexity comes from the interaction between a market and its participants, in other words, the integrated outcome of activities of the entire participants determines the markets trend, while the markets trend affects activities of participants. These interwoven interactions make financial markets keep evolving. Inspired by stochastic recurrent models that successfully capture variability observed in natural sequential data such as speech and video, we propose CLVSA, a hybrid model that consists of stochastic recurrent networks, the sequence-to-sequence architecture, the self- and inter-attention mechanism, and convolutional LSTM units to capture variationally underlying features in raw financial trading data. Our model outperforms basic models, such as convolutional neural network, vanilla LSTM network, and sequence-to-sequence model with attention, based on backtesting results of six futures from January 2010 to December 2017. Our experimental results show that, by introducing an approximate posterior, CLVSA takes advantage of an extra regularizer based on the Kullback-Leibler divergence to prevent itself from overfitting traps.


IEEE Access ◽  
2018 ◽  
Vol 6 ◽  
pp. 19739-19751 ◽  
Author(s):  
Gang Yu ◽  
Jiajun Liu ◽  
Juan Du ◽  
Min Hu ◽  
Vijayan Sugumaran

Author(s):  
Е.П. Трофимов

Предложен алгоритм последовательной обработки данных на основе блочного псевдообращения матриц полного столбцового ранга. Показывается, что формула блочного псевдообращения, лежащая в основе алгоритма, является обобщением одного шага алгоритма Гревиля псевдообращения в невырожденном случае и потому может быть использована для обобщения метода нахождения весов нейросетевой функции LSHDI (linear solutions to higher dimensional interlayer networks), основанного на алгоритме Гревиля. Представленный алгоритм на каждом этапе использует найденные на предыдущих этапах псевдообратные к блокам матрицы и, следовательно, позволяет сократить вычисления не только за счет работы с матрицами меньшего размера, но и за счет повторного использования уже найденной информации. Приводятся примеры применения алгоритма для восстановления искаженных работой фильтра (шума) одномерных сигналов и двумерных сигналов (изображений). Рассматриваются случаи, когда фильтр является статическим, но на практике встречаются ситуации, когда матрица фильтра меняется с течением времени. Описанный алгоритм позволяет непосредственно в процессе получения входного сигнала перестраивать псевдообратную матрицу с учетом изменения одного или нескольких блоков матрицы фильтра, и потому алгоритм может быть использован и в случае зависящих от времени параметров фильтра (шума). Кроме того, как показывают вычислительные эксперименты, формула блочного псевдообращения, на которой основан описываемый алгоритм, хорошо работает и в случае плохо обусловленных матриц, что часто встречается на практике The paper proposes an algorithm for sequential data processing based on block pseudoinverse of full column rank matrixes. It is shown that the block pseudoinverse formula underlying the algorithm is a generalization of one step of the Greville’s pseudoinverse algorithm in the nonsingular case and can also be used as a generalization for finding weights of neural network function in the LSHDI algorithm (linear solutions to higher dimensional interlayer networks). The presented algorithm uses the pseudoinversed matrixes found at each step, and therefore allows one to reduce the computations not only by working with matrixes of smaller size but also by reusing the already found information. Examples of application of the algorithm for signal and image reconstruction are given. The article deals with cases where noise is static but the algorithm is similarly well suited to dynamically changing noises, allowing one to process input data in blocks on the fly, depending on changes. The block pseudoreverse formula, on which the described algorithm is based, works well in the case of ill-conditioned matrixes, which is often encountered in practice


1963 ◽  
Vol 2 (1) ◽  
pp. 37-48 ◽  
Author(s):  
V. P. Turnburke

2021 ◽  
Vol 12 (1) ◽  
Author(s):  
Yoon Ho Jang ◽  
Woohyun Kim ◽  
Jihun Kim ◽  
Kyung Seok Woo ◽  
Hyun Jae Lee ◽  
...  

Abstract Recent advances in physical reservoir computing, which is a type of temporal kernel, have made it possible to perform complicated timing-related tasks using a linear classifier. However, the fixed reservoir dynamics in previous studies have limited application fields. In this study, temporal kernel computing was implemented with a physical kernel that consisted of a W/HfO2/TiN memristor, a capacitor, and a resistor, in which the kernel dynamics could be arbitrarily controlled by changing the circuit parameters. After the capability of the temporal kernel to identify the static MNIST data was proven, the system was adopted to recognize the sequential data, ultrasound (malignancy of lesions) and electrocardiogram (arrhythmia), that had a significantly different time constant (10−7 vs. 1 s). The suggested system feasibly performed the tasks by simply varying the capacitance and resistance. These functionalities demonstrate the high adaptability of the present temporal kernel compared to the previous ones.


Sign in / Sign up

Export Citation Format

Share Document