在 Azure 机器学习工作室(经典)中管理试验运行Manage experiment runs in Azure Machine Learning Studio (classic)

适用于: yes机器学习工作室(经典) no Azure 机器学习APPLIES TO: yesMachine Learning Studio (classic) noAzure Machine Learning

开发预测分析模型是一个迭代过程 - 在修改试验的各种函数和参数时,结果会不断聚合,直到对经过培训的有效模型感到满意为止。Developing a predictive analysis model is an iterative process - as you modify the various functions and parameters of your experiment, your results converge until you are satisfied that you have a trained, effective model. 此过程的关键是跟踪试验参数和配置的各种迭代。Key to this process is tracking the various iterations of your experiment parameters and configurations.

可以在任何时候复查前面运行的试验,以便质询、回顾并最终确认或优化以前的假设。You can review previous runs of your experiments at any time in order to challenge, revisit, and ultimately either confirm or refine previous assumptions. 在运行试验时,机器学习工作室(经典)会保留一份运行历史记录,包括数据集、模块以及端口连接数和参数。When you run an experiment, Machine Learning Studio (classic) keeps a history of the run, including dataset, module, and port connections and parameters. 此历史记录还会捕获结果和运行时信息,如启动和停止时间、日志消息和执行状态。This history also captures results, runtime information such as start and stop times, log messages, and execution status. 可以在任何时候回头查看这些运行,以复查实验和中间结果的时间顺序。You can look back at any of these runs at any time to review the chronology of your experiment and intermediate results. 在创建简单、复杂,甚至建模解决方案的过程中,甚至可以使用上一次运行的试验来启动新一阶段的查询和探索。You can even use a previous run of your experiment to launch into a new phase of inquiry and discovery on your path to creating simple, complex, or even ensemble modeling solutions.

备注

在查看上一次运行的试验时,该版本的试验是锁定的并且不能编辑。When you view a previous run of an experiment, that version of the experiment is locked and can't be edited. 但是,可以通过单击“另存为”并为该副本提供一个新名称来保存一份副本。You can, however, save a copy of it by clicking SAVE AS and providing a new name for the copy. 机器学习工作室(经典)打开新的副本后,就可以编辑和运行该副本了。Machine Learning Studio (classic) opens the new copy, which you can then edit and run. 此实验副本和所有其他试验都位于“试验”列表中。This copy of your experiment is available in the EXPERIMENTS list along with all your other experiments.

查看前一次运行View the prior run

如果有一个你至少运行过一次的试验处于打开状态,可以通过单击属性窗格中的“前一次运行”来查看之前运行的试验。When you have an experiment open that you have run at least once, you can view the preceding run of the experiment by clicking Prior Run in the properties pane.

例如,假设创建了一个试验,并在 11:23、11:42 和 11:55 分别运行了该试验版本。For example, suppose you create an experiment and run versions of it at 11:23, 11:42, and 11:55. 如果打开了最后一次运行的试验 (11:55),则可以单击“前一次运行”打开在 11:42 运行的版本。If you open the last run of the experiment (11:55) and click Prior Run, the version you ran at 11:42 is opened.

查看运行历史记录View the run history

可以通过单击已打开试验中的“查看运行历史记录”来查看该试验的所有以前的版本。You can view all the previous runs of an experiment by clicking View Run History in an open experiment.

例如,假设使用线性回归模块创建了一个试验,并且要观察在更改了试验结果的“学习率”值后的效果。For example, suppose you create an experiment with the Linear Regression module and you want to observe the effect of changing the value of Learning rate on your experiment results. 可以为此参数使用不同的值多次运行此试验,如下所示:You run the experiment multiple times with different values for this parameter, as follows:

学习速率值Learning Rate value 运行开始时间Run start time
0.10.1 9/11/2014 4:18:58 pm9/11/2014 4:18:58 pm
0.20.2 9/11/2014 4:24:33 pm9/11/2014 4:24:33 pm
0.40.4 9/11/2014 4:28:36 pm9/11/2014 4:28:36 pm
0.50.5 9/11/2014 4:33:31 pm9/11/2014 4:33:31 pm

如果单击“查看运行历史记录”,会看到所有这些运行的列表:If you click VIEW RUN HISTORY, you see a list of all these runs:

运行历史记录示例

单击这些运行中的任意一个,以查看在运行时的试验快照。Click any of these runs to view a snapshot of the experiment at the time you ran it. 配置、参数值、注释和结果将全部保留,从而提供试验运行的完整记录。The configuration, parameter values, comments, and results are all preserved to give you a complete record of that run of your experiment.

提示

要记录试验的迭代,可以在每次运行的时候修改标题;可以在属性窗格中更新实验的“摘要”,并且可以为各个模块添加或更新注释以记录所做的更改。To document your iterations of the experiment, you can modify the title each time you run it, you can update the Summary of the experiment in the properties pane, and you can add or update comments on individual modules to record your changes. 每次运行试验都会保存标题、 摘要和模块注释。The title, summary, and module comments are saved with each run of the experiment.

在机器学习工作室(经典)中,位于“试验”选项卡中的实验列表将始终显示实验的最新版本。The list of experiments in the EXPERIMENTS tab in Machine Learning Studio (classic) always displays the latest version of an experiment. 如果打开了前一次运行的试验(使用“前一次运行”或“查看运行历史记录”),可以通过单击“查看运行历史记录”并选择“状态”为“可编辑”的迭代返回草稿版本。If you open a previous run of the experiment (using Prior Run or VIEW RUN HISTORY), you can return to the draft version by clicking VIEW RUN HISTORY and selecting the iteration that has a STATE of Editable.

运行前一次试验Run a previous experiment

在单击“前一次运行”或“查看运行历史记录”打开前一次运行后,可以在只读模式下查看已完成的试验。When you click Prior Run or VIEW RUN HISTORY and open a previous run, you can view a finished experiment in read-only mode.

如果想要从为前一次运行配置迭代的方式开始试验的迭代,可以通过打开运行并单击“另存为”来进行。If you want to begin an iteration of your experiment starting with the way you configured it for a previous run, you can do this by opening the run and clicking SAVE AS. 这会创建一个新的试验,包含新的标题、空的运行历史记录以及前一次运行的所有组件和参数值。This creates a new experiment, with a new title, an empty run history, and all the components and parameter values of the previous run. 此新试验将列在机器学习工作室(经典)主页的“试验”选项卡中,可以修改和运行它,为此试验迭代启动一个新的运行历史记录。This new experiment is listed in the EXPERIMENTS tab in the Machine Learning Studio (classic) home page, and you can modify and run it, initiating a new run history for this iteration of your experiment.

例如,假设有上一部分中所示的试验运行历史记录。For example, suppose you have the experiment run history shown in the previous section. 想要观察在将”学习率“设置为 0.4 以及为”培训时期数”参数尝试不同的值时会发生的情况。You want to observe what happens when you set the Learning rate parameter to 0.4, and try different values for the Number of training epochs parameter.

  1. 单击”查看运行历史记录”并打开在 4:28:36 pm(在此时会参数值设置为 0.4)运行的试验的迭代。Click VIEW RUN HISTORY and open the iteration of the experiment that you ran at 4:28:36 pm (in which you set the parameter value to 0.4).
  2. 单击“另存为”。Click SAVE AS.
  3. 输入新的标题,并单击“确定”复选标记。Enter a new title and click the OK checkmark. 这会创建一份新的试验副本。A new copy of the experiment is created.
  4. 修改“数培训时期”参数。Modify the Number of training epochs parameter.
  5. 单击“运行”。Click RUN.

现在可以继续修改并运行此版本的试验,构建新的运行历史记录以记录所做的工作。You can now continue to modify and run this version of your experiment, building a new run history to record your work.