Compression And Curriculum Strategies For Efficient Learning In Deep Neural Networks