The TensorFlow model represented in TensorBoard contains several basic building blocks of ANNs, whether they are CNNs, RNNs such as LTSM networks, or other forms. All of these models rely on numerical computation in one form or the other: weights, biases, matmul operations, activation functions (logistic sigmoid, ReLU, or others), layers, optimizers, loss functions, and more. All of these tools belong to the field of applied mathematics.
This approach has been around for decades. However, as time goes by, linguists, philosophers, historians, and many more people fascinated by AI will bring new ideas, new programming languages, interfaces, and APIs. Focus on the core concepts, not the tools, and you will always be able to keep up to date and innovate.
Biomimicking humans will inevitably be extended to conceptual representation learning, the fascinating world of our video...
