Schrödinger's FP: Dynamic Adaptation of Floating-Point Containers for Deep Learning Training.
Milos NikolicEnrique Torres-SánchezJiahui WangAli Hadi ZadehMostafa MahmoudAmeer AbdelhadiAndreas MoshovosPublished in: CoRR (2022)
Keyphrases
- floating point
- deep learning
- dynamic adaptation
- deep architectures
- restricted boltzmann machine
- hamilton jacobi
- distance transform
- heat equation
- fixed point
- unsupervised learning
- wave equation
- eikonal equation
- machine learning
- instruction set
- floating point arithmetic
- probabilistic model
- weakly supervised
- mental models
- general purpose
- supervised learning
- computer vision
- object categories
- quality of service
- training examples
- dynamic programming
- evolutionary algorithm
- training data