Tensorboard
"Tensorboard is a flashlight for our Neural Net's black box", as experts say. Tensorboard is a visualization toolikt to understand and inspect your graph.
Tensorboard helps us to visualize our graph to see how the nodes are connected to eachother and how the flow of tensors is propagated inside our network. It also helps us to summerize important parameters (such as wieghts, biases, activations, accuracy, ...) in order to see how each parameter changes in each iteration of the training. We can also visualize the high dimensional tensors in lower dimension spaces using embedding visualization tooklit in tensorflo.
In this tutorial we will explain: