China Food

The whole network starts! Kaggle getting started guide!

Kaggle registration

Step 1:

Kaggle supports two kinds of registration, using Google account registration or email registration. Because kaggle is now acquired by Google, we recommend the first method, that is, register with Google account / email.

If you use other mailboxes to register, it is recommended to use a stable mailbox, such as QQ mailbox. Don’t use the niche email. The email is closed and your kaggle account is lost.

Step 2: fill in your personal information. Here you need a verification code and a network agent (you need to be able to go to Google). You already need a mobile phone number to receive SMS verification code.

Kaggle enters the competition

Step 1:

First log in to the kaggle account, then find the competition you want to participate in, and click in.

Step 2: click [join competition] on the competition page and click [accept] to enter the competition.

Kernel usage and submission

Introduction to kernel Basics

The kaggle kernel is the running environment provided by the kaggle platform to players. It can be regarded as an online notebook. The online nobteook can be used to run its own code (it doesn’t matter whether it is related to the kaggle game or not), and it can also complete the submission of the game through the online notebook.

Click [your work] in the figure below to find your own notebook.

You can find the notebook related to this game in the [Code] interface of a specific kaggle game. These notebooks are related to this competition. Some are used for data analysis, and some are used for modeling.

On the [Code] page of the competition, click the [hotness] button on the right side of the page to switch the sorting method of the notebook, which is sorted by [popularity, likes, comments, creation time, running time, score]. Here you can pay attention to sorting according to [score], and we can easily find the sharing of the highest score.

If you click to enter a specific notebook, similar to the following interface, it is the display state of the notebook, that is, the state after the operation is completed. The specific composition elements are shown in the figure below. At this time, you can click Copy and edit to make a copy of this notebook fork, that is, copy a copy as your own.

Kernel run and commit

If you click [copy and edit] or click [new notebook] on any interface on kaggle, you will enter the editing state of notebook. The editing state of notebook is no different from that of Jupiter notebook. The main interface elements are:

  • Code editing area: run code and support markdown format;
  • Dataset: load the game dataset or model or library installation code
  • Run settings: code run configuration
    • Environment: the running environment, which doesn’t need to be moved;
    • Accelerator:CPU/GPU/TPU
    • GPU quota: GPU weekly running hours
    • Internel: whether it is networked
    • Language: running language, python or R

[note] for the kernel game, it must be submitted from the notebook. At this time, the model prediction code cannot be connected to the network, so it is necessary to complete [library installation], [model loading] and [model prediction] with the Internet off.

If you click the [save version] button in the upper right corner of the notebook in the editing state, the following interface will pop up.

  • Save & run all: will run the notebook from zero;
  • Quick save: save the current code status;

If you click [save version] and select [save & run all], the notebook will enter the running state.

From the clock in the lower left corner of the screen, you can see that the code is running, and running means running. If the queue is in the running queue (running one by one), the previous notebooks may still be running. You can manually stop other notebooks.

If the notebook has been submitted, you can close the notebook editing page. Open the display page of notebook, click version 6 of 6 on the right to find the information about the operation of the submitted code. At this time, notebook is in the running state.

When the notebook operation is completed, it enters the display state again. You can find the output of the code in the output menu on the right side of the notebook that shows the status. Click the submit button to submit, and the notebook enters the submission status.

You can see the submitted notebook in [my submission] of the competition, and you can get scores after the operation is completed.

Dataset usage

Dataset creation and upload

Dataset is a tool provided by kaggle for storing data and models. In the kernel game, the model must be loaded from dataset. The operation of creating dataset is as follows.

You need to open an agent to upload data. You can’t upload data without an agent.

After the dataset is created, you can click the dotted line on the right, and then select [new version] to manage the dataset. You can continue to upload or delete the file.

Dataset usage

Dataset is mainly used to load models and install libraries when the kernel is not connected to the network. In the editing status of notebook, click [add data] on the right to load the dataset.

  • Dataset: own dataset or others’ dataset.
  • Competition data: competition data.
  • Notebook output files: the output of notebook after running can be selected by yourself or others.

When the data set is loaded, you can put the mouse over it and directly copy the path.

Pay attention to the dataset path, otherwise the code will directly report an error.

Kaggle team up

Complete the [team] interface of the competition and enter the names of other teams.

Kaggle anti cheating mechanism

cheating behavior

  • Share code privately without forming a team;
  • Share the dataset privately without forming a team;
  • One machine or one IP login multiple accounts;
  • Others help me submit code or CSV;

Non cheating behavior

  • Run public notebook, submit;
  • Submit the data set published by others;

All the notebooks and datasets that can be seen on kaggle’s official website are public and can be used without cheating.

Recommended reading

Happy birthday: you er hut is three years old!

Kaggle knowledge point: Fundamentals of integrated learning

Seaborn draws 11 histogram

Python crawler travels around the country: Lanzhou station

How to make short videos?

Kaggle dichotomous problem: stroke patient prediction

Kaggle visualization: Black Friday portrait analysis

You er cabin, a warm cabin. The owner of the cabin, who seeks survival with one hand code and enjoys life with one hand spoon, welcomes your presence

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *