dufourc1 / ml_cs433_project2

second project for the course of machine learning CS433 in collaboration with the Laboratory of Computational Systems Biotechnology (SB/SV/STI) LCSB

Hi there 👋

• 🔭 Currently working on combinatorial statistics
• 👨🏼‍🔬 Currently doing a PhD in the Statistical Data Science group in EPFL
• 💬 Interested in machine learning and deep learning applied to graph structured data
• ☕ Surviving mostly on caffeine

🔧 Technologies & Tools

I mostly code in Python and I am starting to discover Julia. I also know my way around R, Matlab, C++, and Java.

Here are a few specialised libraries I have used/am using:

ml_cs433_project2's Issues

features engineering

creation functions to manipulate dataset

categorical variables

write function to cast sigma and mu into categorical variables

plots for final visualization

neural nets

find normalization

need to check the function in charles.py that returns the weights needed in the WLR

write introduction + part about reproduction

Introduction is written (up to our current knowledge of the project).
I will continue writing Material and Methods tomorrow.

lasso, ridge

Already done, need review

split the data wrt to the realization parameter

we should split the data for training and testing in the following way: for each event, we have multiple realization. So split 70% of these realizations (for example) for the training set and 30% for the test set. Do that for all the events and then you get the sets

split the data into train and test and get some goodness of fit measure

reproduction of paper model

reproduction du paper initial :

• find the normalization process
• compute residuals and check the plot
• reproduce the results

run_reproduction.py

reproduce the plots

reproduce the plots so that we can check if the normalization is the good one

svm

Also done in my notebook. Needs review

statsmodel.OLS(X,Y).fit().res

the output is strange, I cannot understand why the residuals given is (n*5) instead of n (n is the number of observations in our dataset)

plot evolution of coefficients wrt to the volume fraction