Introduction to Artificial Intelligence Experiment Management
MissingLink.ai is a powerful deep learning platform that helps data engineers streamline and automate the entire deep learning cycle: data, code, experiments and resources. It eliminates the grunt work and significantly shortens the time it takes to train and deliver effective models.
When it comes to experiment management, integrating the MissingLink SDK to your workflow will immediately provide significant benefits, including:
- Metrics Monitoring: Metrics monitoring that can be easily shared with teammates.
- Graphs: Automatic plotting of graphs for monitored metrics for easy analysis.
- Email Alerts: Get email alerts when events that you are interested in have occurred, such as when an experiment has failed with an exception.
- Experiment Comparison: Compare up to three experiments side-by-side to easily analyze, spot differences, and decide what steps to take to improve the model.
- Confusion Matrix: Automatic plotting of a confusion matrix to help you decide what to do next with your models.
- Simplified Source Tracking: Lets you investigate the code used for running an experiment. Assuming that you use Git, this feature is active by default.
With a little more integration effort, you can enable additional MissingLink capabilities:
- Experiment Hyperparams: Hyperparams annotated on each experiment to help differentiate between different experiments. to help differentiate between different experiments.
- Experiment Custom Metrics: While running an experiment, you can add and monitor custom metrics that are outside the framework, such as the cost of running an experiment.
- Experiment External Metrics: After an experiment has ended, you can add external metrics for monitoring, such as experiments that have met your requirements.
- Graceful Shutdown: You can gracefully shut down an experiment and trigger a new experiment with modified parameters or add a log to your experiment logs, without having an exception raised that triggers an unmanaged shutdown.
- Advanced Source Tracking: You can also track changes that are not committed to source control .
Getting Started with Experiment Management
Now, dive right into the Getting Started Guide for your favorite deep learning framework to learn how to integrate it with our SDK and have your experiments monitored on the MissingLink.ai web dashboard.
- Getting Started for Keras
- Getting Started for TensorFlow with steps
- Getting Started for TensorFlow with batches
- Getting Started for PyCaffe
- Getting Started for PyTorch with steps
- Getting Started for PyTorch with batches
Alternatively, you can try a step-by-step tutorial for setting up Experiment Management in an existing code sample:
- Step-By-Step Tutorial for Integrating with Keras
- Step-By-Step Tutorial for Integrating with TensorFlow
- Step-By-Step Tutorial for Integrating with PyCaffe
- Step-By-Step Tutorial for Integrating with PyTorch
After integrating MissingLink's SDK and seeing some of the benefits in action, you can explore how to integrate the more advanced capabilities of MissingLink's SDK.