https://github.com/innerlee/halfspace
Learning Halfspaces and Neural Networks with Random Initialization
https://github.com/innerlee/halfspace
Last synced: 10 months ago
JSON representation
Learning Halfspaces and Neural Networks with Random Initialization
- Host: GitHub
- URL: https://github.com/innerlee/halfspace
- Owner: innerlee
- License: mit
- Created: 2016-02-25T12:54:47.000Z (almost 10 years ago)
- Default Branch: master
- Last Pushed: 2016-03-01T03:05:33.000Z (almost 10 years ago)
- Last Synced: 2025-02-11T14:46:52.932Z (11 months ago)
- Language: Julia
- Size: 16.6 KB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Halfspace
Implementation of algorithms in "Learning Halfspaces and Neural Networks with Random Initialization"
by Yuchen Zhang, Jason D. Lee, Martin J. Wainwright, and Michael I. Jordan.
Codes written in Julia.
## Required Packages
Install the following packages,
```julia
Pkg.add("JLD")
Pkg.add("StatsBase")
Pkg.add("Convex")
Pkg.add("SCS")
```
## Generate Synthetic Data
First, you need to generate the synthetic data.
Modify code in `run.jl` as
```julia
#include("test/go.jl")
include("test/gendata.jl")
```
Then type
```julia
include("run.jl")
```
in REPL.
Two datasets `set2.jld` and `set5.jld` will be created in `data/` folder.
It does not override existing files.
## Run
Modify code in `run.jl` as
```julia
include("test/go.jl")
#include("test/gendata.jl")
```
Then type
```julia
include("run.jl")
```
in REPL.
Modify `test/go.jl` to switch dataset or change parameters.