mirceamironenco / bayesianrecurrentnn Goto Github PK
View Code? Open in Web Editor NEWImplementation of Bayesian Recurrent Neural Networks by Fortunato et. al
License: MIT License
Implementation of Bayesian Recurrent Neural Networks by Fortunato et. al
License: MIT License
HI, nice work!
I was wondering in bayesian_rnn.py Line 112
why define the prior as pi*math.exp(sigmaone)+(1-pi)math.exp(sigmatwo)
I think reference to your paper, maybe
gaussian1 = torch.distributions.Normal(0,sigma1)
gaussian2 = torch.distributions.Normal(0,sigma2)
prior = pigaussian1+(1-pi)*gaussian2
is correct.
So why this idea?
Good job! Thanks for sharing it!
I have some issues when I use BayesianRNN of your project into my own work. I found that the KL loss is so large which is up to 9.00*10^8 and my prediction loss is 20.00 which is normal, for each time step. So the training process can not continue because the KL loss is not regular. So I dont know the reson why this issue occurs.
I would appreciate it if you can give me some kinds of tips for fixing it or any suggestions.
Hi, I am trying to run this code with tensorflow 1.3.0 and Python3.6
I get a similar error to a separate issue raised:
Traceback (most recent call last):
File "bayesian_rnn.py", line 560, in <module>
run()
File "bayesian_rnn.py", line 510, in run
m = PTBModel(is_training=True, config=config, input_=train_input)
File "bayesian_rnn.py", line 229, in __init__
bbb_bias=config.bbb_bias))
File "bayesian_rnn.py", line 140, in __init__
**self.name = name**
File "/usr/local/lib/python3.6/site-packages/tensorflow/python/keras/engine/base_layer.py", line 1639, in __setattr__
super(Layer, self).__setattr__(name, value)
AttributeError: can't set attribute
The problematic line self.name = name
comes from the BayesianLSTM class. This is the same error as the other issue however I am using Tensorflow 1.3.0 so the issue appears to be around using Python 3.6 instead of Python3.5. Is there a way to amend to code so that it is compatible with Python 3.6.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.