lace 0.9.1

A probabilistic cross-categorization engine
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147

use std::path::PathBuf;

use clap::Parser;
use lace::cc::alg::ColAssignAlg;
use lace::cc::alg::RowAssignAlg;
use lace::cc::feature::FType;
use lace::cc::state::State;
use lace::cc::state::StateGewekeSettings;
use lace::cc::transition::StateTransition;
use lace::geweke::GewekeTester;
use lace::stats::prior_process::PriorProcessType;
use plotly::common::Mode;
use plotly::layout::Layout;
use plotly::Plot;
use plotly::Scatter;
use rand::SeedableRng;
use rand_xoshiro::Xoshiro256Plus;

#[derive(Parser, Debug)]
#[clap(rename_all = "kebab")]
struct Opt {
    #[clap(long, default_value = "gibbs")]
    pub row_alg: RowAssignAlg,
    #[clap(long, default_value = "gibbs")]
    pub col_alg: ColAssignAlg,
    #[clap(long, default_value = "50")]
    pub nrows: usize,
    #[clap(long, default_value = "5")]
    pub ncols: usize,
    #[clap(long)]
    pub no_row_reassign: bool,
    #[clap(long)]
    pub no_col_reassign: bool,
    #[clap(long)]
    pub no_view_alpha: bool,
    #[clap(long)]
    pub no_state_alpha: bool,
    #[clap(long)]
    pub no_priors: bool,
    #[clap(long)]
    pub plot_var: Option<String>,
    #[clap(long, short, default_value = "10000")]
    pub niters: usize,
    #[clap(long)]
    dst: Option<PathBuf>,
}

fn main() {
    let opt = Opt::parse();
    let mut rng = Xoshiro256Plus::from_os_rng();

    // Some of each column type that is supported by Geweke (Labeler cannot be
    // used in Geweke tests)
    let ftypes: Vec<FType> = (0..opt.ncols)
        .map(|i| match i % 3 {
            0 => FType::Continuous,
            1 => FType::Categorical,
            2 => FType::Count,
            _ => unreachable!(),
        })
        .collect();

    // The state's Geweke test settings require the number of rows in the
    // state (50), and the types of each column. Everything else is filled out
    // automatically.
    let mut settings = StateGewekeSettings::new(
        opt.nrows,
        ftypes,
        PriorProcessType::Dirichlet,
        PriorProcessType::Dirichlet,
    );
    let mut transitions: Vec<StateTransition> = Vec::new();

    if !opt.no_col_reassign {
        transitions.push(StateTransition::ColumnAssignment(opt.col_alg));
    }

    if !opt.no_state_alpha {
        transitions.push(StateTransition::StatePriorProcessParams);
    }

    if !opt.no_row_reassign {
        transitions.push(StateTransition::RowAssignment(opt.row_alg));
    }

    if !opt.no_view_alpha {
        transitions.push(StateTransition::ViewPriorProcessParams);
    }

    if !opt.no_priors {
        transitions.push(StateTransition::FeaturePriors);
    }

    transitions.push(StateTransition::ComponentParams);

    settings.transitions = transitions;

    // Initialize a tester given the settings and run.
    let mut geweke: GewekeTester<State> = GewekeTester::new(settings);
    geweke.run(opt.niters, Some(5), &mut rng);

    // Reports the deviation from a perfect correspondence between the
    // forward and posterior CDFs. The best score is zero, the worst possible
    // score is 0.5.
    let res = geweke.result();
    res.report();

    if let Some(ref key) = opt.plot_var {
        use lace::stats::EmpiricalCdf;
        use lace::utils::minmax;
        let (min_f, max_f) = minmax(res.forward.get(key).unwrap());
        let (min_p, max_p) = minmax(res.posterior.get(key).unwrap());
        let x_min = min_f.min(min_p);
        let x_max = max_f.max(max_p);

        let step = (x_max - x_min) / 200.0;
        let xs: Vec<f64> =
            (0..200).map(|i| x_min + (i as f64) * step).collect();

        let cdf_f = EmpiricalCdf::new(res.forward.get(key).unwrap());
        let cdf_p = EmpiricalCdf::new(res.posterior.get(key).unwrap());

        let x_f: Vec<f64> = cdf_f.f(&xs);
        let x_p: Vec<f64> = cdf_p.f(&xs);

        let trace = Scatter::new(x_f, x_p).name("P-P").mode(Mode::Lines);
        let ideal = Scatter::new(vec![0_f64, 1_f64], vec![0_f64, 1_f64])
            .name("Reference")
            .mode(Mode::Lines);

        let mut plot = Plot::new();
        plot.add_trace(ideal);
        plot.add_trace(trace);
        plot.set_layout(Layout::new());
        plot.show();
    }

    if let Some(dst) = opt.dst {
        let f = std::fs::OpenOptions::new()
            .create(true)
            .write(true)
            .truncate(true)
            .open(dst)
            .unwrap();
        serde_json::to_writer(f, &res).unwrap();
    }
}