Another attempt at classifying handwritten digits, but this time using Neural Networks. It gave me percentage error a lot less than what I got using Support Vector Machines. To make this assignment easy, data is provided for hidden layer. This Neural Network contains one input layer with 400 nodes (pixels 20 x 20 of one image), one hidden layer with 25 nodes , and an output layer with 10 nodes (one for each number 0-9). There are 4 files containing data: ps5_data.csv ~ 5000 x 400 matrix of image data, ps5_data-labels.csv ~ 5000 x 1 vector of image labels (10 = "0" label), ps5_theta1.csv ~ 25 x 401 matrix for weights from input layer to hidden layer, and ps5_theta2.csv ~ 10 x 26 matrix for weights from hidden layer to output layer Training data a data set with 5000 handwritten digits and their corresponding labels. Each training example is a 20 pixel by 20 pixel grayscale image of the digit. Each pixel is represented by a number indicating the grayscale intensity at that location. Thus, your neural network will have 400 inputs.
-
Notifications
You must be signed in to change notification settings - Fork 0
Another attempt at classifying handwritten digits (5000 this time), but this time using Neural Networks. It gave me percentage error a lot less than what I got using Support Vector Machines. To make this assignment easy, data is provided for hidden layer. This Neural Network contains one input layer with 400 nodes (pixels 20 x 20 of one image), …
aybidi/Classifying-Hand-Written-Digits---Neural-Networks
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
Another attempt at classifying handwritten digits (5000 this time), but this time using Neural Networks. It gave me percentage error a lot less than what I got using Support Vector Machines. To make this assignment easy, data is provided for hidden layer. This Neural Network contains one input layer with 400 nodes (pixels 20 x 20 of one image), …
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published