jiro_nn 0.8.1

Neural Networks framework with model building & data preprocessing features.
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

use std::{hash::{Hash, Hasher}, path::Path};

use crate::{
    dataset::{Dataset, Feature},
    datatable::DataTable, monitor::TM,
};

use super::{CachedConfig, DataTransformation};

pub struct FeatureExtractorCached {
    pub extracted_feature_config: Box<dyn Fn(&Feature) -> Option<Feature>>,
    pub extract_feature: Box<dyn Fn(&DataTable, &Feature, &Feature) -> DataTable>,
}

impl FeatureExtractorCached {
    pub fn new(
        extracted_feature_config: Box<dyn Fn(&Feature) -> Option<Feature>>,
        extract_feature: Box<dyn Fn(&DataTable, &Feature, &Feature) -> DataTable>,
    ) -> Self {
        Self {
            extracted_feature_config,
            extract_feature,
        }
    }

    fn get_hashed_id(id: &String) -> String {
        let mut hasher = std::collections::hash_map::DefaultHasher::new();
        id.hash(&mut hasher);
        hasher.finish().to_string()
    }

    fn get_cached_feature_file_name(
        &self,
        id: &String,
        working_dir: &str,
        feature: &Feature,
    ) -> String {
        let file_name = Path::new(&working_dir)
            .join("cached")
            .join(format!("{}_{}.csv", Self::get_hashed_id(id), feature.name))
            .to_str()
            .unwrap()
            .to_string();
        file_name
    }

    fn get_cached_feature(
        &self,
        id: &String,
        working_dir: &str,
        feature: &Feature,
    ) -> Option<(DataTable, String)> {
        let file_name = self.get_cached_feature_file_name(id, working_dir, feature);
        if std::path::Path::new(&file_name).exists() {
            let dataset_table = DataTable::from_csv_file(file_name.clone());
            Some((dataset_table.get_column_as_table(&feature.name), file_name))
        } else {
            None
        }
    }

    fn transform_no_cache(
        &mut self,
        mut dataset_table: DataTable,
        feature: &Feature,
        extracted_feature: &Feature,
    ) -> DataTable {
        let old_column = dataset_table.get_column_as_table(&feature.name);
        dataset_table = (self.extract_feature)(&dataset_table, &extracted_feature, feature);
        // if the transformation replaced the old column, we need to add it back
        dataset_table =
            if extracted_feature.name != feature.name && !dataset_table.has_column(&feature.name) {
                dataset_table.append_table_as_column(&old_column)
            } else {
                dataset_table
            };
        dataset_table
    }
}

impl DataTransformation for FeatureExtractorCached {
    fn transform(
        &mut self,
        cached_config: &CachedConfig,
        dataset_config: &Dataset,
        data: &DataTable,
    ) -> (Dataset, DataTable) {
        let mut new_config = dataset_config.clone();
        let mut dataset_table = data.clone();

        if let CachedConfig::Cached { working_dir, .. } = cached_config {
            // create the dataset/cached directory if it does not exist
            std::fs::create_dir_all(format!("{}/cached/", working_dir))
                .expect("Failed to create cache directory");
        }

        for feature in &dataset_config.features {
            if let Some(extracted_feature) = (self.extracted_feature_config)(feature) {
                if let CachedConfig::Cached { id, working_dir } = cached_config {
                    if let Some((cached_data, cachefile_name)) =
                        self.get_cached_feature(&id, working_dir, &extracted_feature)
                    {
                        TM::start("loadcache");
                        dataset_table = if extracted_feature.name == feature.name {
                            dataset_table
                                .drop_column(&feature.name)
                                .append_table_as_column(&cached_data)
                        } else {
                            dataset_table.append_table_as_column(&cached_data)
                        };
                        TM::end_with_message(format!(
                            "Loaded {} from cache {}",
                            extracted_feature.name,
                            cachefile_name
                        ));
                    } else {
                        dataset_table = self.transform_no_cache(dataset_table, feature, &extracted_feature);
                        dataset_table
                            .get_column_as_table(&extracted_feature.name)
                            .to_csv_file(self.get_cached_feature_file_name(
                                &id,
                                working_dir,
                                &extracted_feature,
                            ));
                    }
                } else {
                    dataset_table = self.transform_no_cache(dataset_table, feature, &extracted_feature);
                }

                new_config = if extracted_feature.name == feature.name {
                    new_config.with_replaced_feature(&feature.name, extracted_feature)
                } else {
                    new_config.with_added_feature(extracted_feature)
                };
            }
        }

        (new_config, dataset_table)
    }

    fn reverse_columnswise(&mut self, data: &DataTable) -> DataTable {
        data.clone()
    }

    fn get_name(&self) -> String {
        "".to_string()
    }
}