Neural Turing Machines Tristan Deleu June 23, 2016 @tristandeleu - - PowerPoint PPT Presentation

Neural Turing Machines

Tristan Deleu

@tristandeleu

• June 23, 2016

Deep Learning

The building blocks

Convolutional
 Layer Fully connected
 Layer Recurrent
 Layer

+

Object Recognition Object Detection Image Segmentation Others


• Predictions
• Speech Recognition

Language Processing

Examples

+ =

Object Detection Predictions

+ =

Predictions Speech Recognition

+

Image Segmentation Predictions

Face detection Automatic speech recognition

=

Image segmentation

Examples

+ +

Object Recognition Language Processing Predictions

Sentiment analysis Image captioning Machine translation

= = = +

Language Processing

+

Language Processing Predictions Language Processing

Frameworks

Theano Torch Tensorflow Keras Chainer Neon CNTK MXNet Caffe

Lasagne

Lasagne

Theano + Lasagne

https://github.com/Lasagne/Lasagne/blob/master/examples/mnist.py

Neural Turing Machines

Recurrent Neural Network

ht yt xt ht+1 yt+1 xt+1 yt−1 ht−1 xt−1

LSTMt

LSTMt−1 LSTMt+1

Memory-augmented Networks

BOAT

Neural Network

Boats float on water You can’t sail against the wind Boats do not fly …

?

• Inspired by neuroscience
• Memory-augmented networks: add an external memory to neural networks to act as a knowledge base
• Keep track of intermediate computations — The story to answer the question in QA problems


Memory Networks & Dynamic Memory Networks

Memory-augmented Networks

Memory Networks Dynamic Memory Networks Neural GPU Neural Stack/Queue/DeQue Stack-augmented RNN

Current state Read Operation New state Write 1 1 1

Turing Machine

1 1 1 1 1

q0 q0 q0 q0 q0 q1 q1 q1 q1 · · ·

Neural Turing Machine

1 1 1 1 1

q0

Current state Read Operation New state Write 1 1 1

q0 q0 q0 q0 q1 q1 q1 q1 · · ·

Input Output

?

Heads

1 1 1 1

• wt

Mt

• Turing Machine

Neural Turing Machine

Neural Turing Machine

FFt ht yt rt xt xt FFt+1 ht+1 yt+1 rt+1 xt+1 xt+1 yt−1 ht−1 rt−1 FFt−1 xt−1 xt−1 Mt−1 Mt

• Controller
• Read heads
• Write heads

Neural Turing Machine

ht yt rt xt xt ht+1 yt+1 rt+1 xt+1 xt+1 yt−1 ht−1 rt−1 xt−1 xt−1 Mt−1 Mt

LSTMt LSTMt−1 LSTMt+1

• Controller
• Read heads
• Write heads

Neural Turing Machine

• Memory
• Controller
• Read heads
• Write heads

Input Output

• NTM

Open-source Library

medium.com/snips-ai github.com/snipsco/ntm-lasagne

NTM-Lasagne

Algorithmic Tasks

• Goal: Learn full algorithms only from input/output examples


Generate as much data as we need

• Strong Generalization: Generalize beyond the data the NTM has seen during training


Longer sequences for example

• ?

Input Output

P(X, Y )

?

Copy task

Inputs Outputs EOS

Training

Copy task

Copy task

Copy task

Length 120

Copy task

Length 150

Repeat Copy task

x5 EOS Inputs Outputs

Repeat Copy task

Repeat Copy task

Associative Recall task

Inputs Outputs

Associative Recall task

Associative Recall task

Priority Sort task

bAbI tasks

bAbI tasks

Mary John

bathroom

garden Sandra hallway

Mary John

bathroom

garden Sandra hallway

Mary went to the garden John went to the garden Mary went back to the hallway Sandra journeyed to the bathroom John went to the hallway Mary went to the bathroom

bAbI tasks

Conclusion

• The NTM is able to learn algorithms only from examples
• It shows better generalization performances compared to other recurrent architectures


For example LSTMs

• Fully differentiable structure


Drawback: generalization is still not quite perfect

• New take on Artificial Intelligence


Trying to teach machines things they can do, the same way we would learn them

• Resources
• Theano: http://deeplearning.net/software/theano/
• Lasagne: http://lasagne.readthedocs.io/en/latest/
• NTM-Lasagne: https://github.com/snipsco/ntm-lasagne

@tristandeleu

• June 23, 2016

Thank you

@tristandeleu

• June 23, 2016