Convolution Neural Network

by Xander Cage AI Trainer

What is CNN?

The neural nets exist and in addition to that an image is convoluted, converted in pixel level and studied, converted and a max pooling, this entire thing is known as convolution + pooling layers. A fully connected layer of flattened structure of NumPy array and a hidden layer is then classified into various classes as binary or based on no.of outcomes. In the first layer, we try to understand what each image pixel tries to account for using various filters and then we flatten them and pass it to a neural network, then it goes to a classifier. Want to become a master in Artificial Intelligence check out this Artificial Intelligence Training.

The visual context will go through each and every part of the image and try to understand what is present in each area of the image. The output should be in the form of the class.

Convolutional Neural Network Architecture:

Convolution layer: Here we try to decompose RGB to a multidimensional layer, and apply a filter to each layer. A filter tries to learn all the combinations present in the RGB layer. A strider is used to stride to each matrix in the image. We try to understand this image using a convolution strider.

CNN – Arch :
Given an input image, it goes to convolution+Relu, each area has a 3D, RGB, then it goes to the next pooling layer where it shrinks the max value and this cycle keeps repeating. This is the learning process. We try to classify the values and then we have to apply neural nets and try to figure out what the actual image is. Given that it is a car, softmax gives a value of 0 to 1, the probability of the maximum is identified as the car.
Important aspects of CNN:
The important aspects of CNN are filters, receptive field, stride, padding, pooling, and ReLU layers.

Prepare yourself for the Top Artificial Intelligence Interview Questions And Answers Now!

Filters:
The image is stridden one bit by bit and it ends up till the last window. We have to mostly stride one by one, if we do it double, there might occur a padding problem. The padding size should be nearly half of the slider size. Otherwise, we might miss some information or read other unwanted information. We go for max-pooling always to avoid any such problem. Doing max pooling will cause the matrix size to shrink, from 4×4 to 3×3, and so on.

Training :

Forward Pass
Loss function
Backward pass
Weight update

Learning Rate :
If it is very high, it might not reach Global minima. If it is very low, it will be time-consuming, but the learning is efficient and reached global minima. The formula for learning rate is
where w is the weight, is the initial weight, is the learning rate, and is the change in error factor.
The layer on CNN consists of convolution, ReLU, and pooling together.

Are you interested in learning Artificial Intelligence from experts? Enroll in our AI Course in Bangalore!

Steps to run a CNN :