Intro of Neuron Network

What is neuro network

Neural networks and deep learning currently provide the best solutions to many problems in image recognition, speech recognition.

Structure

1. A neuro network usually contains at least one hidden layer. We can set any amounts of neuron in the hidden layer. In each hidden layer, the input from former layer would be calculated with each neuron. In each neuron, it has weights the same amount as the inputs. Each weight which has been initialized products each inputs, and a bias will be added to the result.
   

2. After the linear calculation, an activation function will be used to the results generated by the neurons.

3. Output. After going through hidden layers, we often use the Softmax function to calculate the possibility of each outputs, the one which has the biggest possibility would be the output
4. Backpropagation. The goal of building up a model is adjusting the parameters in order to make the model to fit inputs well. To this end, we need to use a function to update the parameters like the weights and biases. The gradient decent is in need. We’ll talk about this later.
   
5. Accuracy. We want to evaluate the performance of our model, We use the test data to check the accuracy of our generated model by comparing the predicted label and the exact result.