Faster Convolution Inference Through Using Pre-Calculated Lookup Tables
arXiv:2104.01681
Abstract
Low-cardinality activations permit an algorithm based on fetching the inference values from pre-calculated lookup tables instead of calculating them every time. This algorithm can have extensions, some of which offer abilities beyond those of the currently used algorithms. It also allows for a simpler and more effective CNN-specialized hardware.
11 pages, 7 figures
References in corpus (19)
- Quantizing deep convolutional networks for efficient inference: A whitepaper
- Ternary Weight Networks
- Trained Ternary Quantization
- Speeding-up Convolutional Neural Networks Using Fine-tuned CP-Decomposition
- Training and Inference with Integers in Deep Neural Networks
- Flattened Convolutional Neural Networks for Feedforward Acceleration
- Origami: A 803 GOp/s/W Convolutional Network Accelerator
- Mixed Precision Training of Convolutional Neural Networks using Integer Operations
- Fast Algorithms for Convolutional Neural Networks
- Analysis and Optimization of Convolutional Neural Network Architectures
- EasyQuant: Post-training Quantization via Scale Optimization
- Acceleration of Convolutional Neural Network Using FFT-Based Split Convolutions
- Towards Unified INT8 Training for Convolutional Neural Network
- Complexity Leads to Simplicity: A Consensus Layer V Pyramidal Neuron Can Sustain Interpulse-Interval Coding
- Precision Scaling of Neural Networks for Efficient Audio Processing
- Eyeriss v2: A Flexible Accelerator for Emerging Deep Neural Networks on Mobile Devices
- Depth-wise Decomposition for Accelerating Separable Convolutions in Efficient Convolutional Neural Networks
- Transform Quantization for CNN (Convolutional Neural Network) Compression
- Derivation and Analysis of Fast Bilinear Algorithms for Convolution