Computation graphs, back propagation and neural networks, done simple.
from nn.low import *Provides low-level primitives to create computation graphs, including placeholders, parameters, and basic arithmetic operations.
See nn.examples.linear_regression for the usage of computation graph primitives.
from nn.medium import *Provides higher level operations useful for neural networks, including activations, loss functions, normalization, etc.,
based on basic operation nodes provided by nn.low.
See nn.examples.xor for the usage of medium-level APIs.
from nn.high import *Provides Keras-like high-level abstraction for building and training neural networks.