koba-jon/normal_distribution_dataset

Normal Distribution Dataset

1-dimensional shape dataset generated by random numbers of normal distribution

1. Usage

(1) Original Dataset

Get

$ git clone https://huggingface.co/datasets/koba-jon/normal_distribution_dataset
$ cd normal_distribution_dataset/NormalDistribution
$ ls -l

Hierarchy

• train : training data (100,000 pieces) of 300 dimensions

train
|--0
  |--00000.dat
  |--00001.dat
  |  ...
  |--09999.dat
|--1
|  ...
|--9
  |--90000.dat
  |--90001.dat
  |  ...
  |--99999.dat

• test : test data (1,000 pieces) of 300 dimensions

test
|--000.dat
|--001.dat
|  ...
|--999.dat

(2) Create Dataset

$ sudo apt install python3 python3-pip
$ pip3 install numpy

$ vi scripts/create.sh

--dir : output directory
--num : total number of data
--dim : dimensions on one data
--list : list of mean and standard deviation of normal distribution
--seed : seed of random number

#!/bin/bash

MODE='create'

SCRIPT_DIR=$(cd $(dirname $0); pwd)

python3 ${SCRIPT_DIR}/create.py \
    --dir "./NormalDistribution/${MODE}" \
    --num 100 \
    --dim 300 \
    --list "./list/params.txt" \
    --seed 0

$ sh scripts/create.sh