Shallow Gibbs Double Backpropagation
Project description
shallowgibbs-doublebackpropagation
Double Backpropagation Algorithm -- Implementation with the Shallow Gibbs Model
Double Backpropagation with the Shallow Gibbs Model
Installation
Install using pip
pip install shallowgibbs-doublebackpropagation
Requirements
- Python 3.6 or greater
- scipy
- tensorflow
- pandas
- numpy
- joblib
- scikit-learn
Usage
Import the Shallow Gibbs Double Backpropagation module
import shallowgibbs.doublebackpropagation as SGDBS
You need to load some initial predictions from the Shallow Gibbs Model, or any alike-Structured Model.
The model requires as parameters: the weigths matrix (W), the biases (b), and the response covariance matrix (Sigma).
The model framework backpropagation is updated per observation using:
MSE = \left(y_{i}-\hat{y}_{i}\right)=\|y_{i}-\hat{y}_{e s t, i}\|^{2})
starting from \hat{\psi}_{0}
with the equations:
\begin{gathered}
\hat{\psi}_{1, i} \longleftarrow \hat{\psi}_{0}-\epsilon_{\psi, 0} \frac{\partial M S E\left(y_{i}-\hat{y}_{e s t, i}\right)}{\partial \psi} \\
\hat{\psi}_{t, i} \longleftarrow \hat{\psi}_{t-1, i}-\epsilon_{\psi, t-1} \frac{\partial M S E\left(y_{i}-\hat{y}_{e s t}, i\right)}{\partial \psi}
\end{gathered}
where \psi
is the set of parameters (w,b,Sigma) in our case. They are two additional equations that complete those above
explained in reference [2] and well introduced in [1]. There are about the Training data, and test data predictions update. Please read
reference [2] and the Jupyter Notebook for a guide note of usage and application.
Pypi Project Page
https://pypi.org/project/shallowgibbs-doublebackpropagation/1.0.1/
Notebook Page
https://github.com/kgalahassa/shallowgibbs-doublebackpropagation-notebook
References
[1] Muua, Alejandro, Alahassa, Nonvikan Karl-Augustt. The Shallow Gibbs Network, Double Backpropagation and Differential Machine learning, ScienceOpen Preprints (2021). https://www.scienceopen.com/document?vid=9aab283e-130f-4922-accb-20bef8faff9f
[2] Alejandro Murua, Nonvikan Karl-Augustt Alahassa. Double Back-Propagation and Differential Machine Learning. The Ninth Annual Canadian Statistics Student Conference (CSSC), Jun 2021, Ottawa, Canada. (hal-03265399) https://hal.archives-ouvertes.fr/hal-03265399
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distributions
Hashes for shallowgibbs-doublebackpropagation-1.0.1.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | 40a3ce793266d08a62c174a6b8e6ba0e3edd8038dd8948947e2ee161ae84318d |
|
MD5 | 52983a321ffdda613c8046cf6aab16ec |
|
BLAKE2b-256 | 9436016cddca638277df7575aa19f3ba158c3a3b1100378bc2a509401100daf0 |
Hashes for shallowgibbs_doublebackpropagation-1.0.1-py3.7.egg
Algorithm | Hash digest | |
---|---|---|
SHA256 | bfa9456d6471d10953d516127efc1c124113b2968588b9f9c397aca5f045ebe6 |
|
MD5 | 4d5416af998bb94c9c704cf3ae91af60 |
|
BLAKE2b-256 | eb8cf589338e252280f1c6d434830e286721f3a94eb18bc950dd678b559414b2 |
Hashes for shallowgibbs_doublebackpropagation-1.0.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | c5075d88d999e41d5b6635554719ad0c31693e5a66dc88eb7fdefeb5695f46e6 |
|
MD5 | 868cea9a122cbc3d219365068d86a4a8 |
|
BLAKE2b-256 | a01b8fa036ca048b5852c74f4cc9dafc53d292031f561525d750e4f9b98abd2b |