卷積神經網路
(1) 介紹：
神經網路三大天王NN、CNN和RNN，本文章介紹卷積神經網路(Convolutional Neural Network,CNN)，發明CNN目前為Facebook AI首席科學家被稱為CNN之父，CNN的發明把所有的圖型辨識的演算法幹掉，目前沒有可以比他更準確。當我們利用CNN來訓練辨識貓，CNN會自己找到它有關於貓的特徵，自己去識別．
(2)CNN運作方式
2.1 CNN運作架構
將貓咪像素圖作為輸入神經元，經過Convolution Layer和Pooling Layer擷取圖形特徵值與處理圖形
會先將貓咪圖片轉成一小格一小格的Filter，這些Filter是用來擷取貓身上的特徵值，
2.2 Convolution Layer卷積層
我們會將圖片輸入像素(Pixel)當作輸入的神經元，每個神經元會乘上相同的權重向量(Weight Array)，向量的乘法是做內積，所以由圖片可以看到，將3X3的像素乘以3X3權重值：2*1+1*0+2*1+1*0+4*1+1*0+3*1+1*0+3*1=14．
由左到右，上到下以3X3的區塊遞移計算，最後可以得到一個完整的特徵圖(Feature Map)
我們通常會使用多組Filters(6X6或8X8)去計算出Feature Map，這樣結果會得到原貓咪圖的資料量大很多，如何把這些圖簡化，後來使用一個方法稱為Pooling layer
2.3 Pooling Layer池化層
Pooling Layer層主要的目的是將Filter所得到的Feature Map做簡化，最常用的方法稱為Max-Pooling Layer，將Feature Map以2X2的區塊投票選出最大的值，表示為該區域的特徵值
In this fashion my colleague Wesley Virgin's biography starts in this shocking and controversial video.
回覆刪除As a matter of fact, Wesley was in the military-and shortly after leaving-he found hidden, "mind control" tactics that the government and others used to get whatever they want.
These are the EXACT same secrets tons of famous people (especially those who "became famous out of nothing") and elite business people used to become rich and famous.
You've heard that you use only 10% of your brain.
That's because the majority of your BRAINPOWER is UNCONSCIOUS.
Maybe that thought has even occurred INSIDE OF YOUR very own head... as it did in my good friend Wesley Virgin's head seven years back, while driving an unregistered, garbage bucket of a car with a suspended driver's license and on his banking card.
"I'm very frustrated with living check to check! Why can't I become successful?"
You've taken part in those types of thoughts, ain't it right?
Your own success story is waiting to be written. All you have to do is in YOURSELF.
Learn How To Become A MILLIONAIRE Fast