Utkarsh ERA V6 Part1 Backpropogation

Screen shot of the File

• Nodes: i1, i2
• Description: The input layer is where the network receives its input data. There are two input nodes in this network.

• Weights: w1, w2, w3, w4
• Description: Each connection between the input and hidden layer nodes has an associated weight. These weights adjust the input values as they are passed to the next layer.

• Nodes: h1, h2
• Description: The hidden layer processes the inputs received from the input layer. It combines the weighted inputs and applies a nonlinear sigmoid activation function. The diagram connections (a_h1, a_h2).

• Weights: w5, w6, w7, w8
• Description: These weights modulate the signals being passed to the output layer from the hidden layer.

• Nodes: o1, o2
• Description: The output layer receives the processed signals from the hidden layer and generates the output of the network. Each output node represents a specific output value.

• Activations: a_o1, a_o2
• Description: These denote the activation functions applied at the output nodes, determining the final output of each node.

• Errors: E1, E2
• Description: The network calculates the error for each output node by comparing the actual output with the target output. Error formula suggests L2 Loss

• Node: E_Total
• Description: The total error of the network is calculated by summing the errors from all output nodes. This value is crucial during the training process.

The network is trained using a backpropagation algorithm to adjust the weights based on the calculated error. The goal is to minimize E_Total by iteratively updating the weights to improve the model's predictions.

Loss function reduces linearly

Loss function reduces more steeply

Loss function reduces more steeply compared to 0.2

Loss function reduces more steeply compared to 0.5