An ongoing exploration into neural networks with the end goal of understanding audio analysis at a deeper level. These are living notes — definitions, network types, and the building blocks for eventually applying ML to music data.

intro to neural networks

definitions

features:

weights:

activation function:

• we add positive features and subtract negative features multiplied by their weight
• have to pick a normalization function too (i.e. sigmoid, etc)

neuron:

1. takes the inputs and multiplies them by their weights
2. sums them up
3. applies the activation function to the sum

types

feedforward

• data passes through the different input nodes till it reaches the output node
  • moves in only one direction from the first tier onwards until it reaches the output node
  • this is also known as a front propagated wave which is usually achieved by using a classifying activation function
• the sum of the products of the inputs and their weights are calculated

radial basis

• distance of any point relative to the centre
• 2 layers
  • inner: the features are combined with the radial basis function
  • output of these features is taken into account when calculating the same output in the next time-step