๋ณธ๋ฌธ ๋ฐ”๋กœ๊ฐ€๊ธฐ

728x90

dropout

4/15 ๊ธˆ ์•… ๊ธˆ์š”์ผ!!!!!!!!!!!!!!!!!!!!!!! ๐Ÿ˜‡๐Ÿฅณ ์˜ค๋Š˜์€ MNIST๋ฅผ CNN, Tensorflow 2.x, Colab์œผ๋กœ ๊ตฌํ˜„ํ•œ๋‹ค. Params(weights) = ksize Height ร— ksize Width ร— filter ๊ฐœ์ˆ˜ + b(filter ๊ฐœ์ˆ˜) 1. MNIST by CNN, Tensorflow 2.x, Colab import numpy as np import pandas as pd import tensorflow as tf from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Flatten, Dense from tensorflow.keras.layers import Conv2D, MaxP.. ๋”๋ณด๊ธฐ
4/12 ํ™” ํ™”์š”์ผ! ํ˜„์žฌ์˜ Deep Learning์ด ์–ด๋А ์ •๋„ ํšจ์œจ์„ ๋‚ด๊ธฐ ์‹œ์ž‘ํ•œ ์ด์œ ์— ๋Œ€ํ•ด ๊ณต๋ถ€ํ•œ๋‹ค. Weight์™€ Bias๋ฅผ ๋žœ๋ค ์ดˆ๊ธฐ๊ฐ’์œผ๋กœ ์‚ฌ์šฉํ•˜๋˜ ๊ฒƒ์„ Xievier/He's Initialization์œผ๋กœ ๋Œ€์ฒดํ•˜๊ณ , Vanishing Gradient ํ˜„์ƒ์„ Back-Propagation(ํ–‰๋ ฌ์—ฐ์‚ฐ, ์—ญ์˜ ๋ฐฉํ–ฅ์œผ๋กœ W, b๋ฅผ Update)๊ณผ Activation ํ•จ์ˆ˜๋ฅผ Sigmoid ๋Œ€์‹  ReLU๋ฅผ ์‚ฌ์šฉํ•˜๊ณ , ๊ณ„์‚ฐํ•ด์•ผ ํ•˜๋Š” W, b๋ฅผ Drop-out์œผ๋กœ ์—ฐ์‚ฐ์— ์‚ฌ์šฉ๋˜๋Š” Node๋ฅผ ์ค„์ž„์œผ๋กœ์จ ํ•ด๊ฒฐํ•จ 1. Multinomial Classification by Tensorflow 1.15 ver. import numpy as np import pandas as pd import tensorflow as tf imp.. ๋”๋ณด๊ธฐ

728x90