TensorNodes

TensorNodes allow you to define parts of your model using TensorFlow, and then insert those elements into a Nengo model. TensorNodes work very similarly to a regular Node, except instead of executing arbitrary Python code they execute arbitrary TensorFlow code.

Here is a simple example that uses a TensorNode to compute a sin wave:

import nengo
import nengo_dl
import tensorflow as tf

with nengo.Network() as net:
    node = nengo_dl.TensorNode(lambda t: tf.sin(t))
    p = nengo.Probe(node)

with nengo_dl.Simulator(net) as sim:
    sim.run_steps(1.0)

Note that probing and connecting to TensorNodes works in the same way as regular Nodes.

However, computing sin is something we could do with a regular Node. A more useful application of TensorNodes is in defining network structures that are not easily expressed in Nengo, such as convolutional neural networks. For example, here is a network that applies a convolutional layer to MNIST images:

import nengo
import nengo_dl
import numpy as np
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

class ConvLayer(object):
    def pre_build(self, shape_in, shape_out):
        self.n_mini = shape_in[0] # minibatch size
        self.img_size = int(np.sqrt(shape_in[1])) # image height/width
        self.n_channels = 32 # number of convolutional filters
        self.kernel_size = 3 # convolutional filter size
        self.size_out = shape_out[1] # output dimensionality

    def __call__(self, t, x):
        # reshape input signal to image shape
        image = tf.reshape(x, (self.n_mini, self.img_size, self.img_size, 1))

        # apply convolutional layer
        conv = tf.contrib.layers.conv2d(image, self.n_channels, self.kernel_size)

        # apply dense layer
        dense = tf.contrib.layers.flatten(conv)
        dense = tf.contrib.layers.fully_connected(dense, self.size_out)

        return dense

with nengo.Network() as net:
    # load input data (mnist images)
    mnist = input_data.read_data_sets("MNIST_data/")

    # create node to feed in images
    inp = nengo.Node(nengo.processes.PresentInput(mnist.train.images, 1))

    # create TensorNode to insert the network defined in `ConvLayer`
    tf_node = nengo_dl.TensorNode(ConvLayer(), size_in=28 * 28, size_out=10)

    # create connections to/from TensorNodes, or probe their output, just
    # like a regular Node
    nengo.Connection(inp, tf_node)
    p = nengo.Probe(tf_node)

Note that the above example takes advantage of the pre_build feature of TensorNodes. If the object passed to TensorNode has a pre_build function, NengoDL will call that function once when the model is constructed, and it will pass in the shape of the input and output Tensors. This can be used to define any constants or other operations that don’t need to be executed every simulation timestep.

class nengo_dl.tensor_node.TensorNode(tensor_func, size_in=Default, size_out=Default, label=Default)

Inserts TensorFlow code into a Nengo model. A TensorNode operates in much the same was a a Node, except its inputs and outputs are defined using TensorFlow operations.

The Tensorflow code is defined in a function or callable class (tensor_func). This function accepts the current simulation time as input, or the current simulation time and a Tensor x if node.size_in > 0. x will have shape (sim.minibatch_size, node.size_in), and the function should return a Tensor with shape (sim.minibatch_size, node.size_out). node.size_out will be inferred by calling the function once and checking the output, if it isn’t set when the Node is created.

If tensor_func has a pre_build attribute, that function will be called once when the model is constructed. This can be used to compute any constant values or set up variables – things that don’t need to execute every simulation timestep.

Parameters:

tensor_func : callable

a function that maps node inputs to outputs

size_in : int, optional (Default: None)

the number of elements in the input vector

size_out : int, optional (Default: None)

the number of elements in the output vector (if None, value will be inferred by calling tensor_func)

label : str, optional (Default: None)

a name for the node, used for debugging and visualization