Artificial Intelligence also works on “Cause and effect” i.e. “Latent effect and manifest effect”. If we create a cause or do any action on that “cause” there would be its “effect” that means on the basic of current “actions” there would be a “reaction” due to the nature of the “reaction”, we get rewards and that’s how we learn, so our brain projects those conditions of the environment onto your actions and that’s how you know when you are doing good and when not so much and that’s how we learn.
Let’s discuss Q-Learning Intuition where we have Reinforcement learning in which we have agent who works in a specific environment and perform some actions and as result “State and Rewards” would be given to agent, in case of “State” agent takes its action and would be rewarded accordingly. This would help agent to learn environment and good rewards leads to better state where as bad rewards lead to unfavourable state. Further we have Deep Convolutional Q-Learning and A3C model.
Machine Learning evolves around AI, as data is evolving exponentially hence we need efficient Machine Learning algorithm which has potential to work on real time data. Enterprises are using it to make their operations smarter and more productive.
Deep Learning is most fascinating branch of Artificial Intelligence and Neural Network are most powerful machine learning model based on Deep learning concepts. Deep learning is used for very powerful and intensive tasks like computer vision in medicine. It can be used for variety of purposes from classification and prediction to a business problems or for computer vision like recognition faces and patterns. It can be used to recognise tumors in brain images with the use of Deep Boltzmann machine.
ANN (Artificial Neural Network) is used for regression and classification. Classification problem e.g. prediction of customers who are leaving the bank on the basics of number of given independent parameters like credit score, balance and more.
CNN (Convolutional Neural Network) is used for Computer Vision. Image classification where we provide Input Image to CNN and we get the Image class as output on the basic of Image feature extraction, pooling, and flattering concepts.
RNN (Recurrent Neural Network) is used for time series analysis. RNN is more dynamic and self-learning prediction method, this concept has been taken from the working of brain where output of hidden layers is given back as input to hidden layer which provides Neural Network a self-learning capabilities.
SOM (Self-Organizing Maps) is used to reduce dimensionality. They take the multidimensional dataset where we have lots of Colum and we reduce its dimension. DBM (Deep Boltzmann machine) and AE (Auto Encoders) are used for recommendation system.
As our brain and human existence is evolving so the AI is evolving.