Arcadia AI — Master's Program
EN
Program  /  Semester 1 — AI Foundations
Practicum

Practicum 1: Paper Replication and Reproducible Research Workflow

A first research experience: reproducing a scientific paper from start to finish

MF
Mikhail Fadin

About the course

Students select a real ML paper (NeurIPS/ICML/ICLR-tier) and reproduce its results within a reproducible workflow. The practicum builds skills in critical reading of the literature, experiment organization, and presenting code as a research artifact. The deliverable is a public repository with a notebook, a report on deviations, and a reflection on the original methodology.

What you'll learn

Reproduce the core results of a published ML paper with documented discrepancies
Organize a project to reproducible-research standards (environment, versioning, seed management)
Write a technical report analyzing the original authors' methodological choices

Key topics

Critical reading of ML papers: the anatomy of a Methods section
Reproducible workflow: seeds, environments, data versioning
Git and DVC for research code
Documenting deviations from the original
Presenting results: tables, ablation-study charts
Basic use of compute resources (GPU cluster)
This description was generated automatically and has not yet been reviewed by an instructor — it's a draft for discussion.