探索逻辑回归、神经网络和计算机视觉的互联世界.pptx

1、Exploring the Interconnected World of Logistic Regression,Neural Networks,and Computer VisionLiliang ChenFreddie MAgendaBasics of Logistic Regression Vector addition in 2D Vector addition in 3D defines a line in a two-dimension space vs a hyper-plane in a three-dimension space The unit vector normal

2、 to this plane is The unit vector normal to this line is The unit vector normal to the line/plane has the same direction as the sum of vectors based on individual vectors.Geometric Interpretation of a Logistic RegressionDecision Boundary as a hyperplane in 3D Logistic regression seeks the decision b

3、oundary to perfect linear separate positive and negative points;Decision Boundary as a line in 2D Classification depends on comparing relative distance from the origin to the data points vs.the decision boundary.oGoal of an Activation Function Decision BoundaryActivation function translates a line i

4、nto a nonlinear decision boundary in a 2D space;Logistic regression normally choose sigmoid function as its activation function:Decision Boundary in a Hyper-dimension Space Activation function translates a hyperplane into a complex decision boundary in a higher dimension space*Source:Deep Learning:F

5、eed Forward Neural Networks(FFNNS)-Medium Understand logistic Regression from a One-layer Neural NetworkRewrite using vectorized form,we get while a neural network can have more activation function variation.A logistic regression can be thought of a special case of one-layer neural network.One-layer

6、 neural network can be formulated as which has the same mathematical formula as a logistic regression.Logistic regression uses below sigmoid activation function:Two-layer Neural NetworksZ InputHidden LayerOutputAdding a layer will add the complexity of the networks,but the general forward activation