DSA5208 Project 1: MPI Kernel Ridge Regression

---

Team Members

Name	Email Addess	Student ID
Wong Zi Xin, Avellin	[email protected]	A0225646B
Karthikeyan Vigneshram	[email protected]	A0230109W

---

Deliverables

• In this submission folder, we have included a few files with remarks as follows:
  • housing.tsv: The provided dataset of California housing prices
  • README.md: This current document providing instructions on how to run the MPI pipeline.
  • krr_mpi_script.py: The MPI pipeline with Python binding, with which Kernel Ridge Regression will be implemented in distributed manner within MPI environment
  • summary.pdf: Pdf file that documents the MPI implementation method, HyperParameter tuning details, and best RMSE result.

---

Instructions to Run the MPI Pipeline

Step 1: Install the relevant Python dependencies using pip install

python -m pip install SomePackage
#Replace SomePackage with missing package

Step 2: Install mpi4py

• Run below command in the terminal to install MPI for python if not already done so.

pip install mpi4py

• If on Windows, either Intel MPI or Microsoft MPI is required. Below is the command to install Intel MPI.

pip install impi-rt

Step 3: Run MPI Pipeline

• Run the below command in the terminal.
• The number of processes can be changed by replacing the argument 8 below.
• The argument at the end of the command should either be 0 or 1. 0 indicates to run the kernel ridge regression with the best hyperparameters already tuned previously, 1 indicates to perform hyperparameter tuning (either coarse or fine grain depending on the lines commented out)

mpiexec -n 8 python krr_mpi_script.py 0