Changepoint Detection

The Changepoint Detection workflow monitors RHEED streams in real time and flags changepoints as a growth unfolds: the moments where the detected surface reconstruction changes. Each detected changepoint is matched against reference changepoints you have already labelled elsewhere in the same project, so recurring events inherit consistent, human-meaningful labels. Surfacing these events as they happen lets you respond quickly to unexpected changes in deposition. When enabled, it runs automatically on RHEED data. Detected changepoints are labelled automatically once the sample’s project has labelled reference changepoints. See Setting up reference changepoints.

Automatic detection must be enabled for your deployment. If it is not enabled, new uploads and streams do not generate detections. Contact Atomscale support about availability for your organization. This page describes RHEED changepoint behavior where enabled.

Overview
Technical

Detection Process

For each monitored RHEED stream:

Detect: As the stream arrives, the workflow tracks the surface reconstruction detected in the RHEED pattern. A changepoint fires when the reconstruction changes between states, for example 2x1 to 4x1, or 2x1 to no detectable reconstruction.
Score severity: Each changepoint is assigned a severity (High, Moderate, or Low) based on which transition occurred. By default, every RHEED transition is Moderate.
Label from references: Each new changepoint is compared with the nearest labelled reference changepoints in its project and assigned a label with a confidence score. If the result isn’t confident enough, the changepoint stays Unlabelled until someone labels it.

Setting up reference changepoints

Automatic labels are based on changepoints you have already labelled in the same project. Changepoints are still detected without labelled references, but they stay unlabelled. Setting a project up as the label source takes a few steps.

Choose or create a project

Pick the project that will hold both the reference samples and the sample you want to monitor, or create a new one.

Add labelled reference samples

Add the physical samples from one or more earlier growths to the project, then label the changepoints on their RHEED data items. Reference changepoints are found through the samples in the project, and automatic labeling favors the labels already used on them. Labels themselves are shared across your organization, so you can reuse the same label in any project.

Add the sample you want to monitor

Add the physical sample for the new growth to the same project, so its detected changepoints can be matched against the references.

Stream the growth

Start streaming. Detected changepoints are matched against the project’s references and labelled automatically as they appear.

If the project has no labelled changepoints for this workflow yet, detected changepoints are still flagged, but they remain unlabelled. Label the matching changepoints on samples already in the project to use them as references.

Only labelled changepoints with the same data modality and property are considered as references. Every RHEED changepoint is recorded on the detected surface reconstruction, so any labelled RHEED changepoint in the project can serve as a reference for a new one.

How changepoints appear

Detected changepoints surface in two places in the Monitor view.As bars on the timeseries chart: In the Growth Metrics panel, each changepoint is drawn as a colored vertical band over the chart at the time it occurred. The color encodes severity: red for High, amber for Moderate, and blue for Low. By default, the chart shows only High changepoints, so RHEED changepoints (which default to Moderate) are hidden until you turn them on. To show them, open the chart settings (the gear icon, or right-click the chart and choose Chart Settings) and turn on Moderate and Low under Changepoints, or click Enable All. Instantaneous changepoints are widened to a minimum so they stay visible, and the time axis can be switched between elapsed time and absolute timestamps.In the activity timeline: The Details sidebar’s Activity tab lists every changepoint as a row, ordered through the run, showing its time, a short description, and its label (or Unlabelled) in a badge colored by severity. Select a row to see its Confidence, Deviation, and Assigner. You can filter the timeline by Label, Deviation, and Assigner, relabel an event with Manual Relabel (where you can also edit the label’s description), or request a suggested label with Assistant Relabel. An info card shows which project is supplying the labels and the labels available.

Severity

Severity drives the color of the changepoint in the chart and timeline. For RHEED, it comes from which reconstruction transition occurred:

Severity	Color	Meaning
High	Red	The most severe level. Investigate promptly.
Moderate	Amber	Worth review. RHEED reconstruction changes default to this level.
Low	Blue	A minor event.

RHEED severities come from a built-in table of transitions, so your organization’s saved configuration doesn’t change them.

Detection

Detection runs incrementally as the stream arrives, so changepoints are flagged in near real time rather than only after a run completes. For RHEED, the workflow follows the surface reconstruction classified in each frame, and changepoints are recorded on the detected_rheed_reconstruction property. Two rules shape the raw output:

Debounce: a new reconstruction must persist for several consecutive frames before the transition counts, which suppresses brief flickers.
Transition table: only listed transitions fire, each with its own severity. By default, these are 2x1 to 4x1 and 2x1 to no detectable reconstruction, in either direction, all at Moderate. 1x1 frames are treated as 2x1, so a change between 1x1 and 2x1 never fires, and neither does a change between no detectable reconstruction and 4x1.

Reference-based labeling

When a changepoint is detected, the workflow looks for the closest labelled reference changepoints and uses them to assign a label:

Candidate pool: labelled changepoints on data items in the same project (linked through their physical samples), restricted to the same data modality and property as the new changepoint.
Ranking: candidates are ranked by how close they sit in the run to the new changepoint (the distance between their window centers). Labels confirmed by a person are preferred over unconfirmed automatic suggestions, which keeps earlier guesses from compounding into the taxonomy. Up to five of the closest candidates are used.
Label and confidence: the new changepoint is assigned a label, favoring labels already used on those references, along with a confidence score. Confidence is capped at 90%, so an automatic label never shows as certain.
Confidence gate: results below 50% confidence are discarded, and the changepoint stays unlabelled rather than guessed.
Cold start: if the project has no labelled reference for that modality and property, no label is applied.

The API can also propagate labels across a chosen scope: a single data item, all data items of a physical sample, an entire project, or every data item of the workflow. The app has no control for this.

Severity Assignment

For RHEED, severity comes from the transition table rather than the size of a shift. The table is built in, and organization settings don’t change it.

Configuration

The Changepoint Detection workflow’s Saved Configurations holds one configuration for your organization, with a single setting: Cooldown Period (in ms), the minimum time between consecutive detections.RHEED detection uses a built-in configuration instead, so the saved configuration doesn’t affect RHEED changepoints. When detection is enabled, RHEED detection runs whether or not a saved configuration exists.

Platform

Get Started

Guides

Characterization

Reference

Detection Process

Setting up reference changepoints

How changepoints appear

Severity

Detection

Reference-based labeling

Severity Assignment

Configuration

​Detection Process

​Setting up reference changepoints

​How changepoints appear

​Severity

​Detection

​Reference-based labeling

​Severity Assignment

​Configuration

Detection Process

Setting up reference changepoints

How changepoints appear

Severity

Detection

Reference-based labeling

Severity Assignment

Configuration