Hi Ghulam,
I understand that you want to achieve low error in Neural Networks. You may consider the following ways.
- Normalize the data and consider removing or handling outliers appropriately.
- Increase the network depth by adding more layers which can help in learning more complex patterns.
- Experiment with different activation functions.
- Adding L2 regularization can help prevent overfitting and improve generalization.
I hope this helps!
Thank you.