lightweight

This code replicates the experiments of our paper.

This README file describes the structure of our codes. We will introduce our implementation.

1, Platform:

CPU: 2.40GHz Intel(R) Xeon(R) E5-2680

Ubuntu: 18.04

MATLAB: R2019b

2, Files & Folders

├─baseline
│  ├─kcenter     
│  ├─kmeans      
│  └─kmedian
│              
├─datasets
│  │  data_gen.m
│  │  data_gen_k.m
│  │  data_gen_otl.m
│  │  data_gen_size.m
│          
└─lightweight
    ├─algorithm1
    ├─algorithm2
    ├─algorithm3
    ├─algorithm4
    ├─median_algorithm3
    └─median_algorithm4

3, Interface & Usage

3.1 algorithm1 & algorithm3

[centers,sampNum,k_prime,time] = algorithm1(data,k,z,varargin)
[centers,sampNum,k_prime,time] = algorithm3(data,k,z,varargin)

INPUT:

data is n*d matrix, including n vertices with d attributes; k is the number of clusters; z is the number of outliers;

varargin configure other parameters: eta, delta, ksi, e1, e2, k_prime represent the parameters η, δ, ξ, ε1, ε2, k' respectively, which will affect the value of sampling size, k_prime.

We can also set some parameters directly: sampNum means sampling size, k_prime means the extra centers, RPT means times of repetition, ValdnRatio means Verification set proportion.

OUTPUT:

centers are the clustering centers; sampNum is the sampling size; k_prime is the k'; time is the running time.

EXAMPLE:

[centers,sampNum,k_prime,timer(j)] = algorithm1(data,k,z,'sampNum',S_num,'k_prime',k_prime,'RPT',5,'ValdnRatio',0.02);

3.2 algorithm2 & algorithm4

[centers,sampNum,z_prime,time] = algorithm2(data,k,z,varargin)
[centers,sampNum,z_prime,time] = algorithm4(data,k,z,varargin)

INPUT:

almost the same with algorithm1&algorithm3, except that we set z_prime rather than k_prime in varargin. z_prime represents the parameter z'.

OUTPUT:

almost the same with algorithm1&algorithm3, except that the algorithm outputs z_prime rather than k_prime.

EXAMPLE:

[centers,sampNum,z_prime,timer(j)] = algorithm2(data,k,X.z,'sampNum',S_num,'z_prime',z_prime,'RPT',5,'ValdnRatio',0.2);

4, Details