Quantitative measures to evaluate neural network weight initialization strategies
By Ernesto Zamora Ramos, Masanori Nakakuni, Evangelos A. Yfantis. 2017
21 citations recorded by the source corpus.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Andrés Camero
- Enrique Alba
- Jamal Toutouh
- Thomas Bäck
- Meenal Narkhede
- Prashant P. Bartakke
- Mukul S. Sutaone
- Zwe-Lee Gaing
In Computer Vision and Pattern Recognition
- Andrés Camero
- Enrique Alba
- Jamal Toutouh
- Thomas Bäck
- Meenal Narkhede
- Prashant P. Bartakke
- Mukul S. Sutaone
Citations in the corpus
- Bayesian Neural Architecture Search using A Training-Free Performance Metric (2020, background): “However, in spite of its ability to effectively deal with these issues, the problem still remains open, because the learning process is also affected by the weight initialization strategy [15] and the algorithm parameters [1].”
- Low-Cost Recurrent Neural Network Expected Performance Evaluation (2018, methodology): “Specifically in this study, as a first approach to explore our proposal, we propose to perform the random error sampling using the mean absolute error (MAE) measure [15], initializing the weights using a N(0, 1) distribution [22], and approximating the errors density with a truncated normal distribu”
- A review on weight initialization strategies for neural networks (2021, methodology): “Ramos et al. (2017) have presented an evaluation technique to identify a statistically superior weight initialization scheme among various other weight initialization schemes.”
- A Specialized Evolutionary Strategy Using Mean Absolute Error Random Sampling to Design Recurrent Neural Networks (2019, background): “However, not only the network architecture affects the learning process but also the weight initialization [16] as well as all specific parameters of the optimization algorithm [17].”
- Random error sampling-based recurrent neural network architecture optimization (2019, background): “also the weight initialization [16] and all specific parameters of the optimization algorithm [17].”