Käynnissä

python vowpalwabbit regression with FEARTURES INTERACTIONs

python vowpalwabbit regression for big data files with FEARTURES INTERACTION

1 ridge

2 lasso

3 quantile for both ridge and lasso !

FEATURES BOTH CATEGORICAL AND CONTINUES

IMPORTANT to HAVE INTERACTION BETWEEN CATEGORICAL FEARTURES : SECOND ORDER AND THIRED ORDER

VW should read data from file line by line and train model, not read all file to ram

you need to prove all done correctly for VW

first do VW, then scikit learn for the same data with the same features interactions and show VW performance is not worse than

scikit learn ( no

quantile in scikit learn) and not slower

do features importance and hyperparameters search by VW inner functions

-----------------------------------

seems to be

pyvwis the best

from vowpalwabbit import pyvw

----------------------------------

you choose one data set - not small data has both at least 10 categorical and at least 10 continues features

split data to train and test

[login to view URL]

for example

[login to view URL]

or

[login to view URL]

some ideas , not the best ( not good continues or categorical features )

https://archive.ics.uci.edu/ml/datasets/Online+News+Popularity

or

[login to view URL]

or

[login to view URL]

or

[login to view URL]

all calculations done in vowpalwabbit python including one hot for categorical data (not scikit learn one hot)

code starter

[login to view URL]

[login to view URL]

[login to view URL]

[login to view URL]

[login to view URL]

[login to view URL]

[login to view URL]

[login to view URL]:~:text=Use%20chunksize%20to%20read%20a,be%20read%20in%20per%20chunk.

[login to view URL]

[login to view URL]

[login to view URL]

[login to view URL]

[login to view URL]

vw [login to view URL] -f [login to view URL] –binary –passes 20 -c -q ff –sgd –l1

0.00000001 –l2 0.0000001 –learning_rate 0.5 –loss_function logistic

[login to view URL]

test3 <- c("-t", [login to view URL]("test", "train-sets", "[login to view URL]", package="RVowpalWabbit"),

"-f", [login to view URL](tempdir(), "[login to view URL]"),

"--cache_file", [login to view URL](tempdir(), "[login to view URL]"))

also [login to view URL] many example for VW

maybe >>> from [login to view URL] import DFtoVW

>>> import pandas as pd

>>> df = [login to view URL]({"y": [1], "x": [2]})

>>> conv = DFtoVW.from_colnames(y="y", x="x", df=df)

>>> conv.convert_df()

['1 | x:2']

many code examples

[login to view URL]

some help for all categorical features interactions

[login to view URL]

A basic scenario for one namespace called “a”. You could create quadratic features like this:

vw -d [login to view URL] -q aa

Cubic features must involve three sets:

vw -d [login to view URL] --cubic aaa

Taidot: Python, Keinoäly, Machine Learning (ML)

Näytä lisää: python script copy paste text files, Need a java, php or python script writer (Expert only) — 3, python script to download multiple files from website, python watch directory for new files windows, python watch directory for new files, python monitor directory for new files, python script for editing text files, python poisson regression tutorial, python linear regression, python script to compare two files, python nonlinear regression, python library for parsing log files, python program to merge two files, python split file into multiple files by lines, python split file into multiple files by size, how to run python script on startup - raspberry pi 3, python linear regression numpy, python program to find largest of 3 numbers, convert python code from 2.7 to 3.6 online, python 3.8 features

Tietoa työnantajasta:
( 6 arvostelua ) Toronto, Canada

Projektin tunnus: #31631258

Myönnetty käyttäjälle:

podchasiukdmitro

Thanks for your job posting. I am a Machine Vision Expert. I am very familiar with Deep learning APIs such as Tensorflow, TensorflowLite, TfLearn, Keras, Pytorch, and fastai, mxnet. I have good hands-on working with Ad Lisää

$50 USD 7 päivässä
(0 Arvostelua)
0.0

6 freelanceria on tarjonnut keskimäärin $142 tähän työhön

(199 arvostelua)
6.9
nibeditad007

Hi, Hope you are doing well. I have over 6 years of rich experience in data science and machine learning. I have worked hands on in Python with different datasets for data wrangling, data manipulation, data analysis Lisää

$150 USD 3 päivässä
(11 arvostelua)
4.5
IvanLomakin

Hello. I am a machine learning developer. I have developed several regression project. If you want, I can show you. I am really interested in your project. I want to talk with you about project in more details via cha Lisää

$250 USD 7 päivässä
(2 arvostelua)
2.3
chiranjibpatra

Hi! I am PhD in Computer science. I have read and understood your requirement specifications. I have experience in building matlab/python/c/c++/java based projects for applied sciences. I am confident of taking your Lisää

$200 USD 7 päivässä
(2 arvostelua)
1.8
naveenbelawon

I have 7 years of working experience in machine learning using python. I worked on various regression task with Loss such as lasso. I used VW in various data science project. so I can complete your project on time with Lisää

$100 USD 7 päivässä
(0 arvostelua)
0.0