Recurrent Neural Networks (RNNs) - Intuitive Explanation
Small Correction at 7:08 The 3x4 actually refers to the 3 input attributes and the 4 hidden nodes!
Terminology Clarifications:
2:02 Technically h is the actual vector, and V is a weight matrix, but for this example I described v as the hidden vector entering the model.
6:14 The “hidden” label at the top of the second layer is different from the term “hidden” in the context of RNNs. In this case, hidden just refers to a layer in a neural network that isn’t the input or output layer. This example of a linear layer applies to all the weight matrices used to calculate the hidden vectors for RNNs as well as the output vectors.
----------------------
Get your copy of my free LLMs course: https://www.gptandchill.ai/machine-le...
RNNs or Recurrent Neural Networks were a precursor to the Transformer Neural Network that GPT uses. Today, Google Translate still uses an RNN as part of it's architecture!
