---
title: IU F19 IML Project Report
description: asd
tags: Innopolis University
author: Gleb Petrakov
lang: en-us
---

# Domain Adaptation (SVHN to MNIST) Project Report

Innopolis University, 2019
Introduction to Machine Learning, Fall 2019

Gleb Petrakov
[e-mail]

## Code

The code is available as Kaggle kernel (Apache 2.0):

https://www.kaggle.com/imajou/domain-adaptation-svhn-to-mnist

## Introduction

### About problem

This work is devoted to solving the domain adaptation problem, presented as a course project on Introduction to Machine Learning, Innopolis University.

The problem states two domains:
* SVHN dataset – source domain;
* MNIST dataset – target domain.

While labeled data from source domain are available, the goal is to create a system, able to classify images from target domain without its labels available.

## Concept

### Source
Maximum Classifier Discrepancy for Unsupervised Domain Adaptation
Kuniaki Saito, Kohei Watanabe, Yoshitaka Ushiku, Tatsuya Harada

https://arxiv.org/abs/1712.02560

https://github.com/mil-tokyo/MCD_DA

## Difference from first submission

This is a completely different archtecture, which has nothing to deal with the first submission, except from the framework used.