Description
Introduction to Machine Learning – Artificial Neural Network version 1
1 Introduction
For this assignment you are going to implement an ANN (Artificial Neural Network) to classify the clothing classes in the dataset we provided. You are going to implement various architectures and optimize their hyperparameters. The implementation will be done in Python3, using PyTorch library. After getting the results, you are expected to plot the graphs of the best networks and comment on them.
2 Dataset
The dataset you will be working on is created using Fashion-MNIST Dataset and by adding some color and noise over it. However, the class labels (T-shirt/top, Trouser, Pullover, Dress, Coat, Sandal, Shirt, Sneaker, Bag, Ankle boot) are still the same. The dataset can be downloaded through: http://user. ceng.metu.edu.tr/~artun/ceng499/hw1_data.zip. The dataset has 2 folders: “train” and “test” in which the input images reside. These folders also contain the label files: “labels.txt”. For train split, every line in labels.txt contains image name and its label. However, the labels.txt in test folder only contains image names and the labels for the test split are not given. How you are going to calculate accuracy will be discussed in the later sections. We have not specified the validation split. You are going to create your own validation split from the train split.
3 Test Accuracy Calculation
To emphasize that you cannot use test set during training or hyperparameter optimization, we did not provide the labels for test split. However, you still need to report the test accuracy of your best networks. To calculate your accuracy, you can login the website with ids and passwords in provided on ODTUCLASS:
http://207.154.196.239/
and submit your prediction in the following format:
image_name_1 predicted_label_1 image_name_2 predicted_label_2 . . .
1
