Linear Neural Networks for Classification #

• Single neuron for classification - which category
• Data, linear model with single neuron, sigmoid activation function, and binary cross entropy loss for single neuron for classification (no hidden layers)
• Training using stochastic gradient descent algorithm
• Prediction or inference
• Eg: Breast Cancer Wisconsin (Diagnostic) (UCI) prediction with a single neuron - code implementation from scratch (webinar)
• Multi-class classification - which category
• Data, linear model with multiple output neuron, softmax activation function, and cross entropy loss for multiple neurons for classification (no hidden layers)
• Training using mini-batch stochastic gradient descent algorithm
• Prediction or inference
• Eg: Iris (UCI) prediction with a multiple neuron - code implementation from scratch (webinar)

Reference #

• DNN Module #3 — Linear Neural Networks for Regression. (T1 – Ch 4, T1 - Ch 12)

Home | Deep Learning