dendrites¶
Dendrites aims to be a straightfoward and very easy to use neural network tool for Python3.
Quickstart guide¶
Basic quickstart¶
Here is the basic quickstart, in one piece.
import dendrites
# Create the neural network with 2 layers, input and output layer.
# The Input layer has 2 inputs, and the output layer has 3 outputs.
MyNeuralNet = dendrites.NeuralNetwork(dimensions = (2,3) )
# Add a supervised dataset -- tell the network how to behave
MyNeuralNet.add(input = [0, 1], output = [0, 1, 0])
# Train the network
MyNeuralNet.train()
# Run the network on the dataset we provided earlier, and see the results.
print(MyNeuralNet.run(input = [0, 1]))
# Save the MyNeuralNet in a net.dat file, so we can use it later (without Training it again)
MyNeuralNet.save(location="net.dat")
# Delete MyNeuralNet from memory
del(MyNeuralNet)
# Now lets load the network from the net.dat file we created earlier.
MyLoadedNeuralNet = dendrites.NeuralNetwork()
MyLoadedNeuralNet.load(location="net.dat")
# Lets run the network one more time on the dataset we provided earlier,
# so we can be sure that it is the same network we saved earlier.
print(MyLoadedNeuralNet.run(input = [0, 1]))
Which will produce something like:
[0.057752093363465082, 0.94236022733877001, 0.057788004623543256]
[0.057752093363465082, 0.94236022733877001, 0.057788004623543256]
See also
API documentation to see this code broken down piece by piece, as well as some other features.
Quickstart with logging¶
If you want to use the logging module and see detailed logs from dendrites (recommended), try this:
import dendrites
import logging
# Create the logger
logger = logging.getLogger()
logger.setLevel(logging.DEBUG)
# Create the neural network with 2 layers, input and output layer.
# The Input layer has 2 inputs, and the output layer has 3 outputs.
MyNeuralNet = dendrites.NeuralNetwork(dimensions = (2,3))
# Add a supervised dataset -- tell the network how to behave
MyNeuralNet.add(input = [0, 1], output = [0, 1, 0])
# Train the network
MyNeuralNet.train()
# Run the network on the dataset we provided earlier, and see the results.
logger.info(MyNeuralNet.run(input = [0, 1]))
# Save the MyNeuralNet in a net.dat file, so we can use it later (without Training it again)
MyNeuralNet.save(location="net.dat")
# Delete MyNeuralNet from memory
del(MyNeuralNet)
# Now lets load the network from the net.dat file we created earlier.
MyLoadedNeuralNet = dendrites.NeuralNetwork()
MyLoadedNeuralNet.load(location="net.dat")
# Lets run the network one more time on the dataset we provided earlier,
# so we can be sure that it is the same network we saved earlier.
logger.info(MyLoadedNeuralNet.run(input = [0, 1]))
Which will produce somthing like:
DEBUG:root:[INIT] Initialization started
DEBUG:root:[INIT] Getting network dimensions
DEBUG:root:[INIT] Created a network with dimensions: (2, 3)
DEBUG:root:[INIT] Creating the synapse...
DEBUG:root:[INIT] Created the synapse with 1 elements.
INFO:root:[INIT] Initialization Done
INFO:root:[TRAINING] Training started
DEBUG:root:[TRAINING] Generation = 0, error = 0.6400547667802593
DEBUG:root:[TRAINING] Generation = 10, error = 0.6092681116887855
DEBUG:root:[TRAINING] Generation = 20, error = 0.5802111707589749
# (...)
DEBUG:root:[TRAINING] Generation = 4650, error = 0.010036204401125974
DEBUG:root:[TRAINING] Generation = 4660, error = 0.010012413606938565
INFO:root:[TRAINING] Training done, in 4666 generations, with a final error of 0.009998191544524379
INFO:root:[0.057715475761935281, 0.94223379271320262, 0.057687209685138088]
INFO:root:[SAVE]Saved neural network to file <net.dat>.
DEBUG:root:[INIT] Initialization started
DEBUG:root:[INIT] Getting network dimensions
DEBUG:root:[INIT] Created a network with dimensions: ()
DEBUG:root:[INIT] Creating the synapse...
DEBUG:root:[INIT] Created the synapse with 0 elements.
INFO:root:[INIT] Initialization Done
INFO:root:[LOAD]Loaded neural network from file <net.dat>
INFO:root:[0.057715475761935281, 0.94223379271320262, 0.057687209685138088]
