Skip to main content

Resemblence of Human Neuron to our Artificial Neuron

· One min read
Shaurya Singhal

Source: View original notebook on GitHub

Category: Machine Learning / Learn ML

Introduction​

<img src = 'images/neuron2.png'/>

Resemblence of Human Neuron to our Artificial Neuron

Biological NNArtifical NN
Synaptic Gapweights
DendritesInputs
SomaProcessing Function
AxonActivation Function
Axon Terminals BoutonsOutputs

  • Z is the output of soma
  • a is the output of axon where a = g(Z)
    • g(x) can be sigmoidal function or relu function or any other function.
  • we add bias for activating our neural network (Threshold Thing)

Perceptron (Single Layer Neural Network)​


- A Perceptron is a single layer neural network or say Simplest unitof Neural Network is Perceptron
- It acts a Linear Classifier(Binary classification if activation function if sigmoidal)
- Loss Function : Binary Cross Entropy (-ve of Log Likelihood)
- Optimisation : Gradient Descent/ Stochastic Gradient Descent
- No hidden units
- Input is not counted in layers

<img src = 'images/perceptron.png' width=600/>

Conclusion

Whatever we did (till now) without neural network is basically implementing a perceptron