-
Notifications
You must be signed in to change notification settings - Fork 0
/
README
20 lines (14 loc) · 964 Bytes
/
README
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Aurora
Second Assignment for Dr. Cenek's Machine Learning course at UAA
This project is written in Go, which can be downloaded at https://golang.org/dl/.
It requires the packages go-chart, gonum, and uiprogress.
`go get -u github.com/wcharczuk/go-chart`.
`go get -u gonum.org/v1/gonum/...`
`go get -v github.com/gosuri/uiprogress`
This project can be compiled with `go build -o Aurora`.
Once compiled, it can be ran with `./Aurora prepare` to generate the histogram file, `histograms.dat`. (takes about 1 minute on my machine)
It expects a folder 'images' in the same directory, with subfolders 'aurora' and 'noaurora' containing images to generate histograms for.
Then, it can be ran with `./Aurora run histograms.dat 1000 0.01` to train the perceptron for 1000 epochs with a learning rate of 0.01
and run validation against 20% of the data. (takes about 10 seconds on my machine)
Demonstration of training:
https://asciinema.org/a/9GZrb8fyY0nC2bjXcRnwOV92Y