When access to information on machine learning, neural networks will always encounter picture
Do not understand how this operation is done to identify the object at the beginning, after the completion of the concept of weight and cost function, and it will feel more and more inseparable matrix
Machine learning problems to be solved can be divided into regression and classification, regression: For home prices, there may be a curve showing the relationship between house prices and covers an area of classification problem: data from multiple objects plotted as scatter Figure, there may be a curve to divide the object whether it is garbage. Role of the neural network is to fit these curves, if the true value is B, then the data set is A, the role of the neural network is to find x such a manner as to fit the Ax b. However, the matrix is a linear transformation, resulting in large data sets fit true value has a great difference, it will have a function to activate the input delinearization. After the non-linear neural network has a better fitting ability.