An assortment of exercises and works in progress dealing with machine learning in Javascript.
The data sets used in testing are all publicly available online.
- Regression: Advertising Dataset, courtesy of Gareth James of USC
- KNN (k-Nearest Neighbour): Iris Dataset, courtesy of University of California, Irvine
- Temperature: Randomly generated internally
- MNIST digits: MNIST Image Database
- CIFAR-10 images: CIFAR-10 and CIFAR-100 Datasets
All data sets are expected to reside within the ./src/data folder.
All scripts in ./src folder can be run directly in the node CLI with the following command: node <filename>. Be sure, however, to first run npm install to install the necessary node modules.
The contained scripts can also be run through the given Docker container, by issuing the following: docker-compose run container <filename>. On first run, docker will build the supplied image and install the necessary node modules, before executing the provided filename in node.