Datalore: open beta application for analyzing data in Python
Hi, Habr!
In the ranks of the tools JetBrains replenishment. We are launching the open beta version of Datalore , a smart web environment for analyzing and visualizing data in the Python language.
Machine learning confidently captures the world: data mining algorithms are behind modern commercial developments and research. We have developed an application with which to solve machine learning tasks is easy and pleasant: all the necessary data science tools are available out of the box, and a smart Python code editor makes it easy to analyze data.
')
First , like all JetBrains products, Datalore is a smart code editor: with autocompletions, syntax highlighting and inspections.
The coolest feature of the code editor is intentions. Depending on the task, the relevant intensity appears at the bottom of the edit window - “load dataset”, “train test split” “select model”, etc. Select the desired action, and the code for loading the dataset or model will be automatically generated in the editor cell. And the results will immediately appear in the output window.
Autocompletions and intensities make it easier to work in the code editor. You can stop worrying and focus on data analysis and selection of model parameters.
Secondly , we support incremental calculations: when selecting parameters, replacing one row or adding a cell, you do not need to start the entire workbook from scratch. Datalore automatically starts calculating only those operations that were dependent on edits. Taking into account that all calculations are started automatically, the output window will always contain the most current numbers and tables.
In addition, Datalore gives access to different computing powers depending on the task.
Thirdly, we collected basic data science tools in one web application: from basic libraries for working with data, sklearn machine learning algorithms library and deep learning pytorch algorithms to powerful visualization tools.
The tools are really powerful. Our datalore.plot library is a Python equivalent of the ggplot R language library. The datalore.geo_maps module is designed specifically for creating interactive maps.
If you are just starting to work with data, download one of the pre-installed datasets (Iris, Titanic, MNIST) and play with different models and parameters. And uploading your data is possible through the built-in file manager: it saves the original .csv file and converts it to pandas-dataframe by pressing the corresponding intensity.
Datalore saves the entire analysis process in the cloud. Work begins with a convenient file storage of all workbooks. From here you can share with colleagues a link to the workbook and solve the problem together: Datalore supports simultaneous work of several users.
Changes to the workbook are automatically saved (and no hopelessly lost data). If something went wrong, you can roll back to previous versions of the analysis and track the chronicle of changes using the built-in version control system.
In public beta mode, Datalore is available as a free community plan. In the future, the usage fee will depend on the amount of downloaded data and the capacity used.
Register on datalore.io and load the first dataset!
We are eager to hear about Datalore from all those who often and meticulously work with machine learning tasks: analysts, data scientists, researchers and students. We want to make the most convenient data mining tool and look forward to your feedback. Write to our forum to leave a comment about the product and contact us and other users.
upd 15.02 If you launched a complex algorithm, used more powerful computational instances and received the message “Surcharge limit reached” - please tell us at the forum what problem you were solving, and we will send you a code to activate additional time with instances.
In the ranks of the tools JetBrains replenishment. We are launching the open beta version of Datalore , a smart web environment for analyzing and visualizing data in the Python language.
Machine learning confidently captures the world: data mining algorithms are behind modern commercial developments and research. We have developed an application with which to solve machine learning tasks is easy and pleasant: all the necessary data science tools are available out of the box, and a smart Python code editor makes it easy to analyze data.
')
Datalore features
First , like all JetBrains products, Datalore is a smart code editor: with autocompletions, syntax highlighting and inspections.
The coolest feature of the code editor is intentions. Depending on the task, the relevant intensity appears at the bottom of the edit window - “load dataset”, “train test split” “select model”, etc. Select the desired action, and the code for loading the dataset or model will be automatically generated in the editor cell. And the results will immediately appear in the output window.
Autocompletions and intensities make it easier to work in the code editor. You can stop worrying and focus on data analysis and selection of model parameters.
Secondly , we support incremental calculations: when selecting parameters, replacing one row or adding a cell, you do not need to start the entire workbook from scratch. Datalore automatically starts calculating only those operations that were dependent on edits. Taking into account that all calculations are started automatically, the output window will always contain the most current numbers and tables.
In addition, Datalore gives access to different computing powers depending on the task.
Thirdly, we collected basic data science tools in one web application: from basic libraries for working with data, sklearn machine learning algorithms library and deep learning pytorch algorithms to powerful visualization tools.
The tools are really powerful. Our datalore.plot library is a Python equivalent of the ggplot R language library. The datalore.geo_maps module is designed specifically for creating interactive maps.
If you are just starting to work with data, download one of the pre-installed datasets (Iris, Titanic, MNIST) and play with different models and parameters. And uploading your data is possible through the built-in file manager: it saves the original .csv file and converts it to pandas-dataframe by pressing the corresponding intensity.
Datalore saves the entire analysis process in the cloud. Work begins with a convenient file storage of all workbooks. From here you can share with colleagues a link to the workbook and solve the problem together: Datalore supports simultaneous work of several users.
Changes to the workbook are automatically saved (and no hopelessly lost data). If something went wrong, you can roll back to previous versions of the analysis and track the chronicle of changes using the built-in version control system.
Licensing
In public beta mode, Datalore is available as a free community plan. In the future, the usage fee will depend on the amount of downloaded data and the capacity used.
What's next?
Register on datalore.io and load the first dataset!
We are eager to hear about Datalore from all those who often and meticulously work with machine learning tasks: analysts, data scientists, researchers and students. We want to make the most convenient data mining tool and look forward to your feedback. Write to our forum to leave a comment about the product and contact us and other users.
upd 15.02 If you launched a complex algorithm, used more powerful computational instances and received the message “Surcharge limit reached” - please tell us at the forum what problem you were solving, and we will send you a code to activate additional time with instances.
Source: https://habr.com/ru/post/348696/
All Articles