We propose a novel approach to reduce memory consumption of the backpropagation through time (BPTT) algorithm when training recurrent neural networks (RNNs). Our approach uses dynamic programming to balance a trade-off between caching of intermediate results and recomputation. The algorithm is capable of tightly fitting within almost any user-set memory budget while finding an optimal execution...

Practical Recurrent Learning (PRL) has been proposed as a simple learning algorithm for recurrent neural networks[1][2]. This algorithm enables learning with practical order O(n) of memory capacity and computational cost, which cannot be realized by conventional Back Propagation Through Time (BPTT) or Real Time Recurrent Learning (RTRL). It was shown in the previous work[1] that 3-bit parity pr...

The quality of outdoor spaces in a residential community affects the quality of life of its residents. This paper presents the findings of a study on outdoor thermal comfort and space usage at a residential community in Wuhan, central China, through the monitoring of microclimate conditions, interviews with residents, and recording of occupants’ activities. The data were used to develop a Therm...

A review of the academic discourses on gender and music on the one hand, and on electronic music composition on the other, might elicit the impression that the two fields carry mutually repellent charges.1 Despite the diversity of backgrounds and compositional interests among women electronic composers, their relative obscurity has made theorizing gender difference as a factor in electronic exp...

In the present study, plate impact pressure±shear friction experiments are conducted to provide insight into timeresolved dry sliding characteristics of metal on metal at normal pressures of approximately 1.5 GPa, slip speeds up to 60 m/s and interfacial temperatures as high as 8008C. The plate impact friction experiments represent a signi®cant improvement over conventional dynamic friction exp...

We show that signal ow graph theory provides a simple way to relate two popular algorithms used for adapting dynamic neural networks, real-time backpropagation and backpropagation-through-time. Starting with the ow graph for real-time backpropagation, we use a simple transposition to produce a second graph. The new graph is shown to be interreciprocal with the original and to correspond to the ...