[robotics-worldwide] [news] How to choose the appropriate Machine Learning model ?

sezan92 sezan92 at gmail.com
Fri May 5 23:53:57 PDT 2017

One of the most common questions about Machine Learning right now is the
choice of an ML model for a particular data set . How to choose ? Moreover
which Hyper parameters will be best for my job ? I know the trendiest ML is
Deep Learning, specially for Computer vision. But what about using a less
computationally expensive model for an embedded system ? 

I was working on a project to solve this problem. I have developed a Code
which will show the best Machine Learning Classifiers from the six most
famous Machine Learning Models along with the best hyper parameters for them

I have tested the Code on Five famous data sets- Titanic, MNIST, CIFAR,
Caltech and Machine Learning Donors Dataset from UCI repository. The First
version is available here https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_sezan92_Classifier&d=DwICAg&c=clK7kQUTWtAVEOVIgvi0NU5BOUHhpN0H8p7CSfnc_gI&r=0w3solp5fswiyWF2RL6rSs8MCeFamFEPafDTOhgTfYI&m=ya0r9EDq7DpbLqWcMGFXJdbJVPvTM1KJ8rFtSrvvXwg&s=U8nekCviFaS6_5YeeCOz-Pbpz50bFgkCHMZZSDgszQw&e=  . For MNIST,
CIFAR and Caltech101 data sets I have used images instead of csv files to
show the extraction of features , image pre processing and labelling. Please
use it and give me feedback . 

If you have any question and/or feedback please contact via email
sezan92 at gmail.com . Or The linkedin profile www.linkedin.com/in/sezan92

MD Muhaimin Rahman
Robotics and Computer Vision Engineer

View this message in context: https://urldefense.proofpoint.com/v2/url?u=http-3A__robotics-2Dworldwide.1046236.n5.nabble.com_news-2DHow-2Dto-2Dchoose-2Dthe-2Dappropriate-2DMachine-2DLearning-2Dmodel-2Dtp5716429.html&d=DwICAg&c=clK7kQUTWtAVEOVIgvi0NU5BOUHhpN0H8p7CSfnc_gI&r=0w3solp5fswiyWF2RL6rSs8MCeFamFEPafDTOhgTfYI&m=ya0r9EDq7DpbLqWcMGFXJdbJVPvTM1KJ8rFtSrvvXwg&s=dTkQEn0Uf0yBmk_31HoBKatxB3XoKSmflc3Rifpp-J8&e= 
Sent from the robotics-worldwide mailing list archive at Nabble.com.

More information about the robotics-worldwide mailing list