2018 · 23 citations · 13 references
ComputingComputer EngineeringComputer ScienceMachine LearningNatural Language ProcessingReconfigurable ComputingAccelerator ArchitectureDeep LearningHardware AccelerationPattern RecognitionReconfigurable AcceleratorsSpeech RecognitionAccuracyComputation SpeedData Anomaly DetectionData-driven Anomaly DetectionDeep Learning AccelerationDeep Learning AcceleratorsLong Short-term Memory NetworksRecurrent Neural Networks
Long Short-Term Memory Recurrent neural networks are generally used in speech recognition, machine translation and other fields. And LSTM-RNN also performs well in data anomaly detection. However, Due to the repeatability of LSTM-RNN, general-purpose processors such as CPU and GPGPU cannot efficiently implement LSTM-RNN, most of the existing model optimizations on FPGA are aimed at LSTM cells or large-scale model accelerations that do not require high accuracy (such as speech recognition). For the model of aircraft anomaly detection, which models with short data sampling intervals, high speed requirements and high precision requirements, the accuracy and speed of existing models are insufficient. Therefore, we proposed an FPGA-based LSTM-RNN accelerator to optimize the accuracy and speed of existing models. We achieve the optimization in the computation speed without sacrificing the accuracy, and balance performance and resources utilized in FPGA. The peak performance of our accelerator reaches 13.45 GOP/s, which is superior to other existing methods.
13
Sepp Hochreiter, Jürgen Schmidhuber · Neural Computation · 1997 · 98.1K citations
Klaus Greff, Rupesh K. Srivastava, Jan Koutník et al. · IEEE Transactions on Neural Networks and Learning Systems · 2016 · 6.8K citations · Full text
Learning to Forget: Continual Prediction with LSTM
Felix A. Gers, Jürgen Schmidhuber, Fred Cummins · Neural Computation · 2000 · 5.4K citations