Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/nengo/keras-lmu

Keras implementation of Legendre Memory Units
https://github.com/nengo/keras-lmu

keras legendre lmu lstm nengo recurrent-neural-networks tensorflow

Last synced: 26 days ago
JSON representation

Keras implementation of Legendre Memory Units

Awesome Lists containing this project

README 

        
KerasLMU: Recurrent neural networks using Legendre Memory Units

---------------------------------------------------------------



`Paper `_



This is a Keras-based implementation of the

Legendre Memory Unit (LMU). The LMU is a novel memory cell for recurrent neural

networks that dynamically maintains information across long windows of time using

relatively few resources. It has been shown to perform as well as standard LSTM or

other RNN-based models in a variety of tasks, generally with fewer internal parameters

(see `this paper

`_ for more details). For the Permuted Sequential MNIST (psMNIST) task in particular, it has been demonstrated to outperform the current state-of-the-art results. See the note below for instructions on how to get access to this model.



The LMU is mathematically derived to orthogonalize its continuous-time history – doing

so by solving *d* coupled ordinary differential equations (ODEs), whose phase space

linearly maps onto sliding windows of time via the Legendre polynomials up to degree

*d* − 1 (the example for *d* = 12 is shown below).



.. image:: https://i.imgur.com/Uvl6tj5.png

   :target: https://i.imgur.com/Uvl6tj5.png

   :alt: Legendre polynomials



A single LMU cell expresses the following computational graph, which takes in an input

signal, **x**, and couples a optimal linear memory, **m**, with a nonlinear hidden

state, **h**. By default, this coupling is trained via backpropagation, while the

dynamics of the memory remain fixed.



.. image:: https://i.imgur.com/IJGUVg6.png

   :target: https://i.imgur.com/IJGUVg6.png

   :alt: Computational graph



The discretized **A** and **B** matrices are initialized according to the LMU's

mathematical derivation with respect to some chosen window length, **θ**.

Backpropagation can be used to learn this time-scale, or fine-tune **A** and **B**,

if necessary.



Both the kernels, **W**, and the encoders, **e**, are learned. Intuitively, the kernels

learn to compute nonlinear functions across the memory, while the encoders learn to

project the relevant information into the memory (see `paper

`_ for details).



.. note::



   The ``paper`` branch in the ``lmu`` GitHub repository includes a pre-trained

   Keras/TensorFlow model, located at ``models/psMNIST-standard.hdf5``, which obtains

   a psMNIST result of **97.15%**. Note that the network is using fewer internal

   state-variables and neurons than there are pixels in the input sequence.

   To reproduce the results from `this paper

   `_,

   run the notebooks in the ``experiments`` directory within the ``paper`` branch.



Nengo Examples

--------------



* `LMUs in Nengo (with online learning)

  `_

* `Spiking LMUs in Nengo Loihi (with online learning)

  `_

* `LMUs in NengoDL (reproducing SotA on psMNIST)

  `_



Citation

--------



.. code-block::



   @inproceedings{voelker2019lmu,

     title={Legendre Memory Units: Continuous-Time Representation in Recurrent Neural Networks},

     author={Aaron R. Voelker and Ivana Kaji\'c and Chris Eliasmith},

     booktitle={Advances in Neural Information Processing Systems},

     pages={15544--15553},

     year={2019}

   }