Steboss89

// import rusty_machine for linear regression 
// as well as rust_machine Matrix and Vector as we need them for input 
use rusty_machine; 
use rusty_machine::linalg::Matrix; // surprise, Matrix is in linear algebra
use rusty_machine::linalg::BaseMatrix; // implement traits for matrix
use rusty_machine::linalg::Vector; 
use rusty_machine::learning::lin_reg::LinRegressor;
use rusty_machine::learning::SupModel; // this is extremelly important
// as this implement the traits to user predict and train
//use smartcore::metrics::*;

Steboss89

// Read input data 
let ifile = "FULLPATHTODATASET";
let mut input_data = read_housing_csv(&ifile);

// data preprocessing: train test split 
// define the length of the test data 
let test_chunk_size: f64 = input_data.len() as f64 * 0.3; // we are taking the 30% as test data
// as cast between types 
let test_chunk_size = test_chunk_size.round() as usize;
// split 

Steboss89

// MODEL! 
let mut linearRegression = LinRegressor::default(); 
// train 
linearRegression.train(&x_train_matrix, &y_train_vector);
// predictions
let preds = linearRegression.predict(&x_test_matrix).unwrap();
// convert to matrix both preds and y_test 
let preds_matrix = Matrix::new(test_size, 1, preds); 
let y_preds_matrix = Matrix::new(test_size, 1, y_test);
// compute the mse

Steboss89

use std::vec::Vec; 
use std::env::args; 

mod linear_regression;

fn main() {
    println!("Hello, world!");
    linear_regression::run();
}

Steboss89

pub fn read_housing_csv(filename: impl AsRef<Path>) -> Vec<HousingDataset>

Steboss89

def create_space():
  r""" Function to create the search space"""
  nodes = [8, 16, 32, 64, 128, 256, 512]
  activations = ['relu', 'softmax', 'softplus', 'tanh', 'selu', 'sigmoid']
  # build models 
  known_models = [tf.keras.layers.LSTM, 
                  tf.keras.layer.GRU,
                  tf.keras.layer.Bidirectional]
  
  for i in range(len(nodes)):

Steboss89

def search_strategy():
    r""" Function to combine search space bits together 
    and create a new model"""
    
    # retrieve the output from search space 
    space_output = search_space()
    # from here use a suitable technique to retrieve new models 
    models = MonteCarloSearchTree(space_output)
    # alternatively we could build something like 
    model = Sequential() 

Steboss89

[package]
name = "polars_learning"
version = "0.1.0"
authors = ["Steboss89"]
edition = "2018"

[dependencies]
polars = "0.14.2"

Steboss89

use polars::prelude::*;//{CsvReader, DataType, Field, Result as PolarResult, Schema, DataFrame,};
use polars::prelude::{Result as PolarResult};
use polars::frame::DataFrame;
use std::fs::File;
use std::path::{Path};
// an eye on traits:
use polars::prelude::SerReader;
// If we don't import this one we might get the error CsvReader::new(file) new function not found in CsvReader

pub fn read_csv<P: AsRef<Path>>(path: P) -> PolarResult<DataFrame> {

Steboss89

pub fn deal_with_shape<P: AsRef<Path>>(path: P) -> () {
    /* Example function to retrieve shape info from a dataframe */
    let df = read_csv(&path).unwrap();
    // shape 
    // reming {:#?} otherwise error ^^^^^ `(usize, usize)` cannot be formatted with the default formatter
    let shape = df.shape();
    println!("{:#?}", shape); 
    // schema
    println!("{:#?}", df.schema());
    // dtypes