42. Handwritten Character Classification with AutoML#
42.1. Introduction#
This challenge focuses on the application of automated machine learning. You need to use the auto-sklearn framework to complete the MNIST handwritten character classification.
42.2. Key Points#
MNIST Handwritten Character Classification
Application of the auto-sklearn Framework
42.3. Dataset Introduction#
Digit Recognizer is an introductory machine learning competition on Kaggle. This competition uses the MNIST handwritten character dataset to complete the classification task. The MNIST dataset is very similar to the DIGITS dataset used in the experiment, both consisting of handwritten characters. However, the MNIST samples are larger in size and greater in number.
Exercise 42.1
Challenge: Please read the instructions of the Digit Recognizer competition, use auto-sklearn to complete the competition, and finally try to submit through Kaggle to obtain the ranking results.
Hint: The Kaggle website can be accessed normally on the Chinese mainland, but the registration process may require a “scientific Internet access” environment.
You can download the dataset and complete it locally, or you can complete it online using the Kaggle Notebooks environment provided by Kaggle [recommended]. Finally, you can view the ranking information after submitting the challenge results through Submission Predictions in the upper right corner of the competition page.
When completing locally, the dataset mirror download address:
https://cdn.aibydoing.com/aibydoing/files/digit-recognizer.zip # Copy the link and paste it into the browser to download
The dataset contains 3 files, the explanations are as follows:
├── sample_submission.csv # Sample format for submitting predictions
├── test.csv # Dataset to be predicted in the competition
└── train.csv # Training dataset for the competition
Finally, you can see your ranking in the Leaderboard.
```{note}
It is essential for everyone to complete this open challenge offline by themselves. For the subsequent project challenge competitions in the course, Kaggle will also be used. This challenge can not only help you review the usage of auto-sklearn, but more importantly, familiarize yourself with the use of Kaggle in advance.
Related Links