-
Optical diffractive neural network (ODNN) uses light wave as a computing medium to perform the inference and prediction function of neural network, which has the advantages of high speed, low power consumption, and parallel processing. In this work, an ODNN with only three layers of phase modulation is designed, and a method to improve the recognition performance of ODNN based on the first-order spectral distribution of targets is proposed. Using this method, the parameters of a three-layer ODNN are effectively optimized and the optimal pixel size, diffraction distance, and method for image preprocessing are obtained. The three-layer ODNN designed in this work has a recognition accuracy rate of 95.3% for MNIST (handwritten digit set), compared with the 91.75% accuracy achieved by the five-layer ODNN in the reference (Lin X, Rivenson Y, Yardimci N T, Veli M, Luo Y, Jarrahi M, Ozcan A
2018 Science 361 1004 ). In addition, the system volume is greatly reduced and the system structure is simplified. Combined with the advantages of high speed and low power consumption, it has huge potential applications in the fields such as edge computing in the future.-
Keywords:
- diffractive neural network /
- optical recognition /
- handwritten digits /
- diffractive optical elements
[1] Yann L C, Yoshua B, Geoffrey H 2015 Nature 521 436
Google Scholar
[2] Lin X, Rivenson Y, Yardimci N T, Veli M, Luo Y, Jarrahi M, Ozcan A 2018 Science 361 1004
Google Scholar
[3] Litjens G, Kooi T, Bejnordi B E, Setio A A A, Ciompi F, Ghafoorian M, van der Laak J A W M, van Ginneken B, S´anchez C 2017 Medical Image Analysis 42 60
Google Scholar
[4] Graves A, Mohamed A, Hinton G 2013 IEEE International Conference on Acoustics 2013 6645
Google Scholar
[5] Krizhevsky A, Sutskever I, Hinton G E 2012 In Advances in NIPS 25 1097
Google Scholar
[6] Cho K, Van Merrienboer B, Gulcehre C, Bahdanau D, Bougares F, Schwenk H, Bengio Y 2014 arXiv: 1406.1078[cs. CL]
[7] Silver D, Huang A, Maddison C J, et al. 2016 Nature 529 484
Google Scholar
[8] Shen Y, Harris N C, Skirlo S, et al. 2017 Nat. Photonics 11 441
Google Scholar
[9] Solli D R, Jalali B 2015 Nat. Photonics 9 704
Google Scholar
[10] Psaltis D, Brady D, Gu X G, Lin S 1990 Nature 343 325
Google Scholar
[11] Chang J, Sitzmann V, Dun X, Heidrich W, Wetzstein G 2018 Sci. Rep. 8 123234
[12] Yan T, Wu J, Zhou T, Xie H, Xu F, Fan J, Fang L, Lin X, Dai Q 2019 Phys. Rev. Lett. 123 023901
Google Scholar
[13] Zhou T, Fang L, Yan T, Wu J, Li Y, Fan J, Wu H, Lin X, Dai Q 2020 Photonics Research 8 940
Google Scholar
[14] Mengu D, Luo Y, Rivenson Y, Ozcan A 2020 IEEE J. Sel. Top. Quant. 26 3700114
Google Scholar
[15] Li J, Mengu D, Luo Y, Riven Y, Ozcan A 2019 Advanced Photonics 1 046001
[16] Sun Y, Dong M, Yu M, Xia J, Zhang X, Bai Y, Lu L, Zhu L 2021 International J. Optics 2021 6667495
[17] Lécun Y, Bottou L, Bengio Y, Haffner P 1998 Proceedings of the IEEE 86 2278
Google Scholar
[18] Xu P, Hong C, Cheng G, Zhou L, Sun Z 2015 Optics Express 23 6773
Google Scholar
[19] 徐海东, 徐平, 亓森林, 梁钱, 张旭琳, 黄洁锋, 黄海漩 2010 光学学报 30 1002
Google Scholar
Xu H D, Xu P, Qi S L, Liang Q, Zhang X L, Huang J F, Huang H X 2010 Acta Opt. Sin. 30 1002
Google Scholar
[20] Yang T, Chen M, Xiao Y, Xu H, Xu P 2020 IEEE Photonics J. 12 7800310
Google Scholar
[21] Xu P, Yang T, Kai L, Li X, Long S, Huang H 2018 IEEE Photonics J. 10 7801210
Google Scholar
[22] Ouabida E, Essadique A, Bouzid A 2017 Expert Systems with Applications 71 383
Google Scholar
[23] 胡文刚, 王永仲, 华文深 2006 光学技术 32 179
Google Scholar
Hu W G, Wang Y Z, Hua W S 2006 Optical Technique 32 179
Google Scholar
[24] 苏显渝, 吕乃光, 陈家璧 2010 信息光学原理 (第一版) (北京: 电子工业出版社) 第34—40页
Su X Y, Lü N G, Chen J B 2010 Principles of Information Optics (Vol. 1) (Beijing: Publishing House of Electronics Industry) pp34–40 (in Chinese)
[25] Zhang Y, Wallace B 2015 arXiv: 1510.03820 [cs. CL]
-
图 1 (a) 深度神经网络; (b) 经典的范德鲁特光学相关器
Figure 1. (a) Deep neural network; (b) the classic Vander Lugt correlator.
图 2 三层DOE衍射神经网络 (a) 光路示意图; (b) 程序流程图
Figure 2. Three-layer DOE ODNN: (a) Optical path; (b) program flow chart.
图 3 接收面的区域划分及演示说明 (a) 10个阿拉伯数字所在的区域划分; (b) 当输入图像为数字8, 代表数字8的区域所在的能量最大
Figure 3. Area division and demonstration description of the receiving surface: (a) Area division of the 10 Arabic numerals are located; (b) when the input image is the digit 8, the area representing the digit 8 has the maximum energy.
图 4 手写数字及对应的频谱幅度图
Figure 4. Handwritten digits and the corresponding spectral map.
图 5 不同像素大小的图像和频谱幅度谱
Figure 5. Image and amplitude spectrum of different pixel sizes.
图 6 ODNN的系统分辨率对识别率的影响
Figure 6. Impact of ODNN’s resolution on recognition rate.
图 7 衍射距离D对信息传递的影响
Figure 7. Effect of diffraction distance on information transmission.
图 8 衍射距离D对ODNN识别率的影响
Figure 8. Influence of diffraction distance D on the recognition rate of ODNN.
图 9 三层ODNN的训练结果
Figure 9. Training results of three-layer ODNN.
图 10 ODNN的 (a) 输入和 (b) 输出图例; (c) DOE量化后的仿真结果; (d) 量化前及量化后的DOE
Figure 10. Input (a) and output (b) pattern examples of ODNN; (c) simulation result when the DOEs are quantified; (d) DOEs and quantified DOEs.
-
[1] Yann L C, Yoshua B, Geoffrey H 2015 Nature 521 436
Google Scholar
[2] Lin X, Rivenson Y, Yardimci N T, Veli M, Luo Y, Jarrahi M, Ozcan A 2018 Science 361 1004
Google Scholar
[3] Litjens G, Kooi T, Bejnordi B E, Setio A A A, Ciompi F, Ghafoorian M, van der Laak J A W M, van Ginneken B, S´anchez C 2017 Medical Image Analysis 42 60
Google Scholar
[4] Graves A, Mohamed A, Hinton G 2013 IEEE International Conference on Acoustics 2013 6645
Google Scholar
[5] Krizhevsky A, Sutskever I, Hinton G E 2012 In Advances in NIPS 25 1097
Google Scholar
[6] Cho K, Van Merrienboer B, Gulcehre C, Bahdanau D, Bougares F, Schwenk H, Bengio Y 2014 arXiv: 1406.1078[cs. CL]
[7] Silver D, Huang A, Maddison C J, et al. 2016 Nature 529 484
Google Scholar
[8] Shen Y, Harris N C, Skirlo S, et al. 2017 Nat. Photonics 11 441
Google Scholar
[9] Solli D R, Jalali B 2015 Nat. Photonics 9 704
Google Scholar
[10] Psaltis D, Brady D, Gu X G, Lin S 1990 Nature 343 325
Google Scholar
[11] Chang J, Sitzmann V, Dun X, Heidrich W, Wetzstein G 2018 Sci. Rep. 8 123234
[12] Yan T, Wu J, Zhou T, Xie H, Xu F, Fan J, Fang L, Lin X, Dai Q 2019 Phys. Rev. Lett. 123 023901
Google Scholar
[13] Zhou T, Fang L, Yan T, Wu J, Li Y, Fan J, Wu H, Lin X, Dai Q 2020 Photonics Research 8 940
Google Scholar
[14] Mengu D, Luo Y, Rivenson Y, Ozcan A 2020 IEEE J. Sel. Top. Quant. 26 3700114
Google Scholar
[15] Li J, Mengu D, Luo Y, Riven Y, Ozcan A 2019 Advanced Photonics 1 046001
[16] Sun Y, Dong M, Yu M, Xia J, Zhang X, Bai Y, Lu L, Zhu L 2021 International J. Optics 2021 6667495
[17] Lécun Y, Bottou L, Bengio Y, Haffner P 1998 Proceedings of the IEEE 86 2278
Google Scholar
[18] Xu P, Hong C, Cheng G, Zhou L, Sun Z 2015 Optics Express 23 6773
Google Scholar
[19] 徐海东, 徐平, 亓森林, 梁钱, 张旭琳, 黄洁锋, 黄海漩 2010 光学学报 30 1002
Google Scholar
Xu H D, Xu P, Qi S L, Liang Q, Zhang X L, Huang J F, Huang H X 2010 Acta Opt. Sin. 30 1002
Google Scholar
[20] Yang T, Chen M, Xiao Y, Xu H, Xu P 2020 IEEE Photonics J. 12 7800310
Google Scholar
[21] Xu P, Yang T, Kai L, Li X, Long S, Huang H 2018 IEEE Photonics J. 10 7801210
Google Scholar
[22] Ouabida E, Essadique A, Bouzid A 2017 Expert Systems with Applications 71 383
Google Scholar
[23] 胡文刚, 王永仲, 华文深 2006 光学技术 32 179
Google Scholar
Hu W G, Wang Y Z, Hua W S 2006 Optical Technique 32 179
Google Scholar
[24] 苏显渝, 吕乃光, 陈家璧 2010 信息光学原理 (第一版) (北京: 电子工业出版社) 第34—40页
Su X Y, Lü N G, Chen J B 2010 Principles of Information Optics (Vol. 1) (Beijing: Publishing House of Electronics Industry) pp34–40 (in Chinese)
[25] Zhang Y, Wallace B 2015 arXiv: 1510.03820 [cs. CL]
DownLoad:
Catalog
Metrics
- Abstract views: 3735
- PDF Downloads: 178
- Cited By: 0
