jaakik/ml-project-1

Higgs Boson Challenge EPFL CS-433

You can download the challenge data following the instructions in AiCrowd

Data:

Create a data folder and place the test.csv and train.csv files following this projce folde structure :

root ── main 
      ├─ data ─── test.csv
      │       ├─ train.csv
      │                    

Utils:

Please checkout data_utils.py and train_utils.py for all the helper functions that were involved in data processing and training.

Implementations:

Please checkout implementations.py for the implementation of the cost functions that were used in this project

Training & Prediciton

Run run.py to generate the predictions using ridge regression with fine tuned hyper parameters.

Requirements

• Python >= 3.8
• numpy

ml-project-1's Issues

Train function

Setting up Overleaf for report

Normalize / Standardize

1. Provide helpers function to clean the data ie remove undesired values like the -999
2. remove outliers
3. normalize / standardize the data

Function 3&4

fix PCA

PCA

• Implement PCA numpy

Function 1 & 2

Run Function 5 & 6

1. Run functions 5 & 6 on clean data
2. provide some plots for the loss and other metrics of interest
3. make submission to aicrowd

function 5 & 6

Pipeline

• Preprocessing method that encapsulate all data exploration / cleaning / normalization
• Split train validation
• Batching

Read paper

• For physical insights about the data

Data pipeline

Integrate the dataset for training

Polynomial expansion of features