nitarshan / bayes-by-backprop Goto Github PK
View Code? Open in Web Editor NEWPyTorch implementation of "Weight Uncertainty in Neural Networks"
Home Page: https://www.nitarshan.com/bayes-by-backprop/
PyTorch implementation of "Weight Uncertainty in Neural Networks"
Home Page: https://www.nitarshan.com/bayes-by-backprop/
The return statement should be:
return self.l1.log_variational_posterior
+ self.l2.log_variational_posterior
+ self.l3.log_variational_posterior
instead of the current :
return self.l1.log_variational_posterior
+ self.l2.log_variational_posterior
+ self.l2.log_variational_posterior
Currently, the log posterior of l2 is being counted twice and the log posterior of l3 is not being taken into account at all.
Hi, shouldn't pi be a parameter rather than a fixed value during training? Because you don't know what the real distribution of weight is. Thanks.
Is .backward() doing the rep trick automatically?
Hi, I can figure out the role of self.training, please help me
Hi, in Class BayesianLinear, the weight is sampled from a unimodal Gaussian distribution
self.weight = Gaussian(self.weight_mu, self.weight_rho)
, and the prior of weight is sampled from a mixture Gaussian
self.weight_prior = ScaleMixtureGaussian(PI, SIGMA_1, SIGMA_2)
But is it possible to use a single Gaussian distribution to approximate mixture Gaussian distribution theoretically? It seems meaningless...
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.