-
Notifications
You must be signed in to change notification settings - Fork 1
Home
In this small project, I used/implemented some programming concepts in order to classify mnist data. Among those concepts are the following: First: ensemble a set of models into a single model. It is a grouping of several relatively weak models together to produce a strong model, those models are neural networks of the convoluted type with a few layers.
Also, the resulting model of assembly was applied to the stored data mnist, and through it I was able to reduce the overfitting that we obtained in the previous project to a relatively large proportion.
The accuracy level of the training data was 99.11%, while the accuracy level of the test data was 99.1%, and these results are very good, aren't they? The second concept: Functional API in Keras. In programming models, I used the concept of API, this method has many strong advantages, including that we can call an entire model as a layer of another model.
I did the application in an easy way.