AI from Great Learning.md

• Linear vs Non-Linear Classifier
• White vs Black box
• Perceptron, it's types and failure
  • MCNeuron has no concepts of weight so it gives equal weight to all nodes. It has no bias layers, only input and output and step function
• Artificial Neural Network
  • Dense
  • Partial
  • Number of layers has no rule
  • By default Gradient Decent is used for error correction
  • Nothing but Matrix Algebra and a sequence of Matrix Operations
  • If number of Neurons in a layer is higher then the number of attributes then it's projction in higher dimention space and thus more distinct classifications can be made
  • weights are modified to minimizing the error | example using gradient decent