psamnn: PyTorch Neural Net classification on the PSAM spam database

Bart Massey

This repository contains a couple of things:

• personal.csv contains instances from the PSAM spam corpus of some years ago. The instances consist of a name, a class (1 for spam, 0 for ham), and a vector of features obtained via big-bag-of-words and SpamAssassin analyses.

• Python code for a simple PyTorch Neural Net to classify the instances.

Running

You will need PyTorch installed. Easiest is to say

pip3 install -r requirements.txt

To run on the "personal" corpus with 10-way cross-validation and a neighbor distance of 5, say

python3 knn.py personal.csv 10 5

The output will consist of the accuracy for each cross-validation split.