Click or drag to resize
Creating a Simple Experiment
[This topic is pre-release documentation and is subject to change in future releases. Blank topics are included as placeholders.]

In this procedure, we'll create a simple analytical model consisting of a single dataset connected to a single statistical module. We will then run the experiment, edit it, and run it again.

Tip Tip

For more general information about how to create and iterate on an experiment, see Creating and Editing an Experiment.

We will follow these general steps:

  1. Create a new experiment canvas

  2. Add a dataset

  3. Add a module

  4. Connect the dataset and module

  5. Run the experiment

  6. Modify the experiment and run it again

Tip Tip

ML Studio saves your work every 5 seconds, but at any time you can also save your work by clicking SAVE below the experiment canvas. This saves the current working state of your experiment, and you can return to the draft later by finding it in the list in the EXPERIMENTS tab of ML Studio.

You can save a copy of your experiment with a new name by clicking SAVE AS.

Note that if you navigate to another page in the browser or close the browser before the experiment is run or saved, you will lose your work.

Our experiment will look like this:

Figure 1: Sample Experiment

Create a new experiment canvas

  1. At the bottom of the ML Studio window, click +NEW and then select EXPERIMENT. A new, blank experiment canvas is displayed.

  2. Select the current experiment title ("Untitled") at the top of the canvas and change it to something meaningful, such as "Sample Draft 1".

Add a dataset

  1. To the left of the ML Studio window, find the dataset, Iris Two Class Data. You can find it by scrolling through the list (datasets are listed first, followed by modules), or by typing the dataset name in the Search box at the top of the list.

  2. Using your mouse, drag and drop the dataset onto the experiment canvas.

    Notice that the dataset has a triangle at the bottom labeled dataset - this is the output port of the dataset. To view the data in the dataset as text, right-click the output port and select View. You can also view the data by selecting the dataset and clicking View dataset in the pane to the right of the canvas.

    Note Note

    Depending on the format of a dataset, you may not be able to view the data this way directly in the browser. In that case, you will be prompted to download the data and view it in a separate application.

Add a module

  1. To the left of the ML Studio window, find the module, Descriptive Statistics. You can find it by scrolling through the list, or by typing the module name in the Search box at the top of the list.

  2. Using your mouse, drag and drop the module onto the experiment canvas, somewhere below the Iris Two Class Data dataset you added above.

    Notice that the module has a triangle on top - this is the input port of the module - and another triangle on the bottom - this is the output port.

    Note Note

    After dragging the module to the canvas, a red exclamation mark appears on the module. Hover your mouse over the mark and you will see the message, "Input port dataset is unconnected." This alert will go away once we connect the module to the dataset.

Connect the dataset and module

  • Left-click the dataset output port and, while holding the mouse button down, drag the mouse to the input port of the module, then release the mouse button. This connects the output of the Iris Two Class Data dataset to the input of the Descriptive Statistics module, creating a simple data flow.

    If you had other modules in your experiment, you could connect the output of the dataset to multiple modules by repeating this drag-and-drop procedure. You could also link modules together by connecting the output of one module to the input of another in the same way.

Run the experiment

  1. Below the experiment canvas, click RUN.

    The RUN command submits your experiment to ML Studio to be processed. A clock icon appears on the module indicating that the experiment has been scheduled. If we had added more than one module the clock icon would appear on each.

    When the experiment has finished processing, each clock icon changes to a check mark. If necessary, you can click REFRESH below the canvas to update the status. If there had been an error during processing, the clock icon would have changed to an exclamation mark - you can display the error message by hovering over the exclamation mark.

  2. Right-click the output port of the Descriptive Statistics module to view the results.

    Note Note

    Note that not all results can be viewed as text directly in the browser. In those cases you would be prompted to download the data to view it in a separate application.

Modify the experiment and run it again

  1. To iterate on your model, you can add or remove datasets and modules or change module parameters. At any time you can click SAVE below the canvas to save your changes to the experiment.

  2. When you have finished, click RUN to run it again.

    When you run the experiment again, ML Studio only executes the portion of the experiment that you have changed. If you leave a module unchanged, and the data flowing to the module has not changed, then ML Studio does not execute that module again.

  3. To preserve an iteration, click SAVE AS below the canvas and enter a meaningful name for the experiment title. This saves a copy of your experiment which you then can modify and run as a new iteration. To discard any changes you have made since the last time you ran the experiment, click DISCARD CHANGES.

    You can find all the experiments that you have saved in the EXPERIMENTS tab of ML Studio.