Comparison of MNIST classification with/without autoencoder using Logistic Regression, Linear SVM, Random Forest, and Neural Network. In this experiment the output from encoder is used as input to the classification algorithm.
Data | Logistic Regression | Linear SVM | Random Forest | Neural Network |
---|---|---|---|---|
Original | 0.9201 | 0.9183 | 0.9471 | 0.9832 |
Encoded | 0.8746 | 0.8804 | 0.9182 | 0.9799 |
- Python 2
- Keras
- matplotlib
- numpy
- tensorflow