A memo when executing the deep learning sample code created from scratch with Google Colaboratory
Introduction This is the content of ** "Deep Learning from scratch-The theory and implementation of deep learning learned with Python" published by O'REILY ** when executing the sample code in the Google Colaboratory environment.
--Get sample data of books from GitHub --About how to mount data (Introduction of connection method because it cannot be used just by uploading data to My Drive) --About image display (Since the image was not displayed in Google Colaboratory even if the sample code was written, an alternative method was introduced)
Chapter 3 Neural Network 3.6.1 ** MNIST dataset ** 3.7.1 ** Neural network inference processing **
It will be a memo when you execute the above contents. (P72〜P75)
Data acquisition
** Get data from GitHub. **(download)
https://github.com/oreilly-japan/deep-learning-from-scratch
Folder name = ** deep-learning-from-scratch-master **
About uploading and mounting data
1. Upload the folder in My Drive ** Google Drive **.
2. Create a new Google Colaboratory.
3. Mount the drive. (Make the data available in the notebook)
--Click the icon to start the process.
--Connect to Google Drive
Move directory
You need to change the current directory before entering the first code.
Import of load_mnist function
Run mnist_show.py
――It is possible to omit the code up to the 6th line. (Already executed when importing load_mnist)
--Number of the first training image = ** 5 **
--Number of one-dimensional arrays = ** 784, **
--Retransformed from the one-dimensional array to the original shape = ** 28, 28 **
--The training image that should have been displayed is not displayed. = ** It is stored in img but is not displayed. ** **
Image display
--I tried another display method to check if it is really ** 5 **.
--Import matplotlib and display the image.
――We succeeded in displaying the first training image.
Run neuralnet_mnist.py
--Correct recognition accuracy = ** 0.9352 (93.52%). ** **
Finally
I just ran the sample code in Google Colaboratory, but I spent a lot of time.
I posted it for the first time so that people who are worried about similar content can use it as much as possible.
Recommended Posts
Run mnist_show.py
――It is possible to omit the code up to the 6th line. (Already executed when importing load_mnist)
--Number of the first training image = ** 5 **
--Number of one-dimensional arrays = ** 784, **
--Retransformed from the one-dimensional array to the original shape = ** 28, 28 **
--The training image that should have been displayed is not displayed. = ** It is stored in img but is not displayed. ** **
Image display
--I tried another display method to check if it is really ** 5 **.
--Import matplotlib and display the image.
――We succeeded in displaying the first training image.
Run neuralnet_mnist.py
--Correct recognition accuracy = ** 0.9352 (93.52%). ** **
Finally
I just ran the sample code in Google Colaboratory, but I spent a lot of time.
I posted it for the first time so that people who are worried about similar content can use it as much as possible.
Recommended Posts
Run neuralnet_mnist.py
--Correct recognition accuracy = ** 0.9352 (93.52%). ** **
Finally
I just ran the sample code in Google Colaboratory, but I spent a lot of time.
I posted it for the first time so that people who are worried about similar content can use it as much as possible.
Recommended Posts
Recommended Posts