Skip to content
master
Switch branches/tags
Go to file
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
src
 
 
 
 
 
 
 
 

Divide and Conquer Networks (DiCoNet)

figgeneral Code accompanying Divide and Conquer Networks

DiCoNet Summary

Model

eq1

Weak supervision Loss

eq1

Gradients computation

eq1

Reproduce Experiments

Prerequisites

  • Python 3.6.1 + some traditional libraries
  • Computer with Linux or OSX
  • PyTorch
  • For training, an NVIDIA GPU is needed. CPU not supported.

Convex Hull

Baseline

python code/ConvexHull2d/main.py --path [experiment folder] --path_dataset [dataset folder]

Without split computational regularization

python code/ConvexHull2d/main.py --path [experiment folder] --path_dataset [dataset folder] --dynamic

Add split computational regularization

python code/ConvexHull2d/main.py --path [experiment folder] --path_dataset [dataset folder] --dynamic --regularize_split

K-means

Check the parse arguments section at the top of the kmeans.py file to change the default arguments.

python code/K-means/main.py --path [experiment folder] --path_dataset [dataset folder] --dataset ["GM"/"CIFAR"]

Knapsack

Compile the knapsack solver for the creation of the dataset.

g++ src/Knapsack/solver.cc -O2 -o src/Knapsack/solver

Train the model. Check the parse arguments section at the top of the knapsack.py file to change the default arguments.

python src/Knapsack/knapsack.py --dataset_path [dataset folder] --solver_path src/Knapsack/solver --logs_path [experiment folder]

About

No description, website, or topics provided.

Resources

License

Releases

No releases published

Packages

No packages published