serenade 0.3.1

Serenade: Session-based recommender system
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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175

use std::cmp::min;
use rand::SeedableRng as _;
use serenade::objective;
use serenade::config::AppConfig;
use std::io::{stdout, Write};

use indicatif::ProgressBar;

extern crate csv;
use csv::Writer;

pub fn convert_string_to_vec_i32(s: String) -> Vec<i32> {
    return s
        .replace(&['[', ']'][..], "")
        .chars().filter(|c| !c.is_whitespace())
        .collect::<String>()
        .split(",")
        .map(|s| s.parse().unwrap())
        .collect();
}

fn main() -> anyhow::Result<()>{
    // get params from config file
    let config_path = std::env::args().
        nth(1).
        expect("Config file not specified!");
    let config = AppConfig::new(config_path);
    let training_data_path = config.hyperparam.training_data_path;
    let test_data_path = config.hyperparam.test_data_path;
    let validation_data_path = config.hyperparam.validation_data_path;
    let num_iterations = config.hyperparam.num_iterations;
    let save_records = config.hyperparam.save_records;
    let out_path = config.hyperparam.out_path;
    let enable_business_logic = config.hyperparam.enable_business_logic;
    let n_most_recent_sessions_range = convert_string_to_vec_i32(
        config.hyperparam.n_most_recent_sessions_range);
    let neighborhood_size_k_range = convert_string_to_vec_i32(
        config.hyperparam.neighborhood_size_k_range);
    let last_items_in_session_range = convert_string_to_vec_i32(
        config.hyperparam.last_items_in_session_range);
    let idf_weighting_range = convert_string_to_vec_i32(
        config.hyperparam.idf_weighting_range);

    // Progress bar
    let pb = ProgressBar::new(num_iterations as u64);

    let mut wtr = Writer::from_path(out_path)?;
    if save_records {
        // csv writer for storing all values of the whole procedure
        wtr.write_record(&[
            "iteration",
            "n_most_recent_sessions",
            "neighborhood_size_k",
            "last_items_in_session",
            "idf_weighting",
            "MRR@20"
        ])?;
    }
    /*
    Each instance of TpeOptimizer tries to search out the value
    which could minimize/maximize the evaluation result
    for such hyperparameter.
    */
    let mut optim0 =
            // n most recent sessions
            tpe::TpeOptimizer::new(tpe::parzen_estimator(), tpe::range(n_most_recent_sessions_range.first().unwrap().clone() as f64, n_most_recent_sessions_range.last().unwrap().clone() as f64)?);

    let mut optim1 =
            // neighbourhood size k
            tpe::TpeOptimizer::new(tpe::parzen_estimator(), tpe::range(neighborhood_size_k_range.first().unwrap().clone() as f64, neighborhood_size_k_range.last().unwrap().clone() as f64)?);

    let mut optim2 =
            // last items from session
            tpe::TpeOptimizer::new(tpe::parzen_estimator(), tpe::range(last_items_in_session_range.first().unwrap().clone() as f64, last_items_in_session_range.last().unwrap().clone() as f64)?);

    let mut optim3 =
        // last items from session
        tpe::TpeOptimizer::new(tpe::parzen_estimator(), tpe::range(idf_weighting_range.first().unwrap().clone() as f64, idf_weighting_range.last().unwrap().clone() as f64)?);

    println!("===============================================================");
    println!("===           START HYPER PARAMETER OPTIMIZATION           ====");
    println!("===============================================================");

    // mutable variables
    let mut best_value = std::f64::NEG_INFINITY;
    let mut rng = rand::rngs::StdRng::from_seed(Default::default());

    // optimization loop for num_iterations
    for i in 0..num_iterations {
        // increment progress bar
        pb.inc(1);
        // ask() gets the next value of the optimization target
        // hyperparameter to be evaluated
        let n_most_recent_sessions = optim0.ask(&mut rng)?;
        let neighborhood_size_k = optim1.ask(&mut rng)?;
        let last_items_in_session = optim2.ask(&mut rng)?;
        let idf_weighting = optim3.ask(&mut rng)?.floor();
        // get the result of the object function
        // with current combination of hyperparameters
        let v = objective::objective(
            training_data_path.clone(),
            validation_data_path.clone(),
            n_most_recent_sessions as i32,
            neighborhood_size_k as i32,
            last_items_in_session as i32,
            idf_weighting,
            enable_business_logic
        );

        if save_records {
            // Save current values
            wtr.write_record(&[
                i.to_string(),
                n_most_recent_sessions.to_string(),
                neighborhood_size_k.to_string(),
                last_items_in_session.to_string(),
                idf_weighting.to_string(),
                v.to_string()
            ])?;
        }

        // Tells the evaluation result of a hyperparameter
        // value to the optimizer
        optim0.tell(n_most_recent_sessions, v)?;
        optim1.tell(neighborhood_size_k, v)?;
        optim2.tell(last_items_in_session, v)?;
        optim3.tell(idf_weighting, v)?;

        // update current best_value
        best_value = best_value.max(v);

    }

    println!("Considering {} iterations for hyper parameter optimization...", num_iterations);

    let n_most_recent_sessions = optim0.trials().into_iter()
        .find(|(_value, score)| score == &best_value).map(|(value, _score)| value as i32).unwrap();
    let neighborhood_size_k = optim1.trials().into_iter()
        .find(|(_value, score)| score == &best_value).map(|(value, _score)| value as i32).unwrap();
    let neighborhood_size_k = min(neighborhood_size_k, n_most_recent_sessions);
    let last_items_in_session = optim2.trials().into_iter()
        .find(|(_value, score)| score == &best_value).map(|(value, _score)| value as i32).unwrap();
    let idf_weighting = optim3.trials().into_iter()
        .find(|(_value, score)| score == &best_value).map(|(value, _score)| value.floor()).unwrap();

    let test_score = objective::objective(
        training_data_path.clone(),
        test_data_path.clone(),
        n_most_recent_sessions,
        neighborhood_size_k,
        last_items_in_session,
        idf_weighting ,
        enable_business_logic,
    );
    let evaluation_length = 20;
    println!("===============================================================");
    println!("===          HYPER PARAMETER OPTIMIZATION RESULTS          ====");
    println!("===============================================================");
    println!("MRR@{} for validation data: {:.4}", evaluation_length, best_value);
    println!("MRR@{} for test data: {:.4}", evaluation_length, test_score);

    println!("enabled business_logic for evaluation:{}", enable_business_logic);
    println!("best hyperparameter values:");
    println!("n_most_recent_sessions:{}", n_most_recent_sessions);
    println!("neighborhood_size_k:{}", neighborhood_size_k);
    println!("idf_weighting:{}", idf_weighting);
    println!("last_items_in_session:{}", last_items_in_session);

    println!("HPO done");
    println!("HPO done");
    stdout().flush()?;
    wtr.flush()?;

    Ok(())
}