On Deep Learning
One of the main advantages is that it learns features in the hidden layer.But good feature engineering can work as well. Activation functions also allow non linear learning. They can result in neural network being able to learn any function.
Other advantages include generalization using embeddings in NLP. RNN etc. allow a bit more flexibility for long sequences.
Beyond that I think traditional algos can also work well. My system will guarantee beat deep learning for NLIDB. 100 percent guarantee. For chatbot, deep learning can be more useful.
Other advantages include generalization using embeddings in NLP. RNN etc. allow a bit more flexibility for long sequences.
Beyond that I think traditional algos can also work well. My system will guarantee beat deep learning for NLIDB. 100 percent guarantee. For chatbot, deep learning can be more useful.
There is no doubt it is the best unsupervised system.
Comments
Post a Comment