講演情報
[18p-P06-11]Magneto-Optical diffractive deep neural Networks by Monte Carlo Method
〇FatimaZahra Chafi1, Hotaka Sakaguchi1, Hirofumi Nonaka2, Hiroyuki Awano3, Takayuki Ishibashi1 (1.Nagaoka Univ. Tech., 2.Aichi Inst. Tech., 3.Toyota Tech. Inst.)
キーワード：
MO-D2NN、Monte Carlo Method、Learning
Deep Neural Networks have been surprisingly developed and implemented in various applications. However, as the DNNs have become more complicated, the high-speed computing and the energy consumption are much challenging. Recently, magneto-optical diffractive deep neural network (MO-D2NN) device has been proposed [1,2]. Here, we used Monte Carlo method to train MO-D2NN to recognize the handwritten digits, the distance between the hidden layers are optimized. The model of this calculation contains of two hidden layers with the distance between the input layer and the first layer, the distance between the hidden layers and the distance from the second layer to the output are considered as: 2.0, 3.5, and 2.0 mm, respectively. The incident linearly polarized light is used as a light source with a wavelength of 532 nm. MNIST dataset has been chosen as input data for the classification of the handwritten digits. Magnetic domain patterns in the hidden layers, which were rondomnly prepared, were rewritten using the MCM with 1000 steps for each epoch (epochs=30), for 1,000 inputs in parallel process. Figure 1 shows the recorded magnetic domain patterns at the size of 2x2 μm2 after learning by: a. MCM for 2 hidden layers with the distances 2.0, 3,5, 2.0 mm, b. MCM for two hidden layers with the distances 2.0, 2.0, 1.0 mm [3], and c. Gradient descent backpropagation five hidden layers with distances 0.5 mm [1]. It is clearly observed that the magnetic domain is successfully determined using MCM, and the accuracy is imporved accordingly. Figure 2 presents the accuracy and the loss evolution against the steps of MCM. This confirms the reliability of the Monte Carlo method into the training process of the MO-D2NN
コメント
コメントの閲覧・投稿にはログインが必要です。ログイン