What are differences between logistic regression and a Perceptron. I know perceptrons are used for building NNs but looking just at the single perceptron identity I’m not able to figure out the difference.
- Both perform weighted sum on inputs to reach an intermediate output.
- Both perform transformation on this intermediate output to reach final output. In logistic regression, we use sigmoid function for this purpose.
- In both architectures, we update the weights based on error. We apply gradient descent in both the architectures.
- Both are fundamentally used to classify data points which are linearly separable.
Can i say perceptron is logistic regression with some cosmetic changes?