aKountable

Safe, accountable AI in the Blockchain!
by saferAI

Installation

pip install aKountable

Usage

First, say STOP! to people cheating on AI audits.

Start getting Tierion credentials for free. You will need a username, password and api keys. This is what we use to talk with the Blockchain.

Create a Keras model and an aKountable object

model = Sequential()
model.add(Dense(512, activation='relu', input_shape=(784,)))
model.add(Dropout(0.2))
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.2))
model.add(Dense(10, activation='softmax'))

aK = aKountable(model, username, password, api_key)
aK.model.summary()
aK.model.compile(loss='categorical_crossentropy',
                 optimizer=RMSprop(),
                 metrics=['accuracy'])

Train your model to solve hard AI problems

history = aK.model.fit(x_train, y_train,
                       batch_size=batch_size,
                       epochs=epochs,
                       verbose=1,
                       validation_data=(x_test, y_test))

Stamp your model using the Blockchain and get a receipt

aK.stamp()
aK.receipt()

Be aKountable!

aK.validate()

If successful, get yourself saferAI aKountable badge! See full example here.

Documentation

CisD

ToDO

• Add functionality to save aKountable receipt together with Keras model
• Use logging instead of print.
• Extend aKountable. For all AI, Machine Learning and Neural Network library, add model description AND model weights to a hash and use an API that can talk with the Blockchain.
• aKountable as a Service: aKaaS
• Work with regulators to make bring aKountable and safer AI to everyone.