Chainer – Knowledge and References

Explore chapters and articles related to this topic

Implementation

Published in Seyedeh Leili Mirtaheri, Reza Shahbazian, Machine Learning Theory to Applications, 2022

Seyedeh Leili Mirtaheri, Reza Shahbazian

Chainer is a Python-based, independent open-source framework for Deep Learning models. Its major developers work at Preferred Networks, Inc., a Machine Learning startup with engineers mostly from the University of Tokyo. A wide range of Deep Learning models, including RNN, CNN, reinforcement learning, and variational autoencoders, is provided by Chainer.

IoTouch: whole-body tactile sensing technology toward the tele-touch

View Article

Journal Information

Published in Advanced Robotics, 2021

Van Anh Ho, Shotaro Nakayama

The environment used for the experiments (PC with Intel CPU corei5-7300u 2.6 GHz, Windows 10, Python 3.7.6) is close to the reported specifications of [7]. Programming language is Python for handling multiple tasks at one language, such as building and training network models for machine learning, image pre-processing, sensor deformation simulation, and data visualization. Chainer is used as the library for deep learning. Using the prepared dataset, training was performed on a desktop PC equipped with a GPU, with a mini-batch size set to 1000. The learning rate, which is related to the accuracy and progress of the training, was set using the library's automatic setting function. Here, the weights updated as the error back propagation was calculated in the learning process, which is called ‘step’ and the maximum number of learning steps is set to 1000. In addition, the learning progress was determined and the system is set to terminate if the evaluation value is not updated for 30 continuous steps.