Convolutionally Implementation of Sliding window Help Need

Hi Mentor,

@thearkamitra
@arosacastillo
@AmmarMohanna
@XpRienzo
@reinoudbosch
@chrismoroney39
@paulinpaloalto

We are unable to understand the concept in the lecture Convolutional Implementation of Sliding Window. Please help us on this, we are struck into this video.

Statement 1 at 7:45 minute, It turns out that this blue 1 by 1 by 4 subset gives
you the result of running in the upper left hand corner 14 by 14 image. This upper right 1 by 1 by 4 volume gives you the upper right result. The lower left gives you the results of implementing the convnet on the lower left 14 by 14 region. And the lower right 1 by 1 by 4 volume gives you the same result as running the convnet on the lower right 14 by 14 medium

Doubt in statement 1 : We are unable to get intuition behind this statement. How it reduced four times convnet forward propagation computation to one forward propagation. Because above statement telling like convnet is running on the left , right , top , bottom corner of the 14 by 14 image then this implies convnet applied four times right ie convnet applying across all four different positions but why so one time forward propagation saying in the lecture?

Statement 2 at 8:53 : Instead, it combines all four into one form of computation and shares a lot of the computation in the regions of image that are common

Doubt 2: Can u please explain with example of this statement (shares a lot of the computation in the regions of image that are common)…how computation shared ?

Statement 3 at 9:54 Because of the max pooling up too that this corresponds to running your neural network with a stride of two on the original image. what does it mean sir ?

@XpRienzo can you please sir help on this ? Im struck and unable to understand

@XpRienzo Sir please kindly do needful when u get time…we can understand u are busy but we are unable to procees next video…

can you please help on this ?

Did you find any solution?

This is a year old thread. The best idea is to watch the video again. Prof Ng explains things in his always excellent way. If after doing that you still have questions, please create your own new thread and ask your own specific questions.

I agree with Paul: the topic is well explained by Prof. Ng in the course material. You might want to watch the video twice and pause + google or ask if a specific term / context is not clear.

Personally, it helped me that I got to know convolution in university from signal and system theory in the course of signal processing and control engineering.

Therefore, maybe this thread and the linked video helps to get familiar with the underlying concept, @Anbu, @Hossain_Piash: How to Calculate the Convolution? - #2 by Christian_Simonis

Please let me know if you have any questions!

Best regards
Christian