Enum BoundaryHandlingMethod

Namespace

AiDotNet.Enums

Assembly

AiDotNet.dll

Specifies how to handle boundaries when processing data that extends beyond the available range.

public enum BoundaryHandlingMethod

Fields

Periodic = 0

Treats the data as if it repeats infinitely in all directions.

For Beginners: With Periodic boundary handling, the data is treated as if it wraps around from one edge to the opposite edge, like a loop.

Imagine a photo where walking off the right edge makes you reappear at the left edge, or walking off the bottom makes you reappear at the top.

Example: For a 1D array [A,B,C,D,E]:

• If you need a value one position to the left of A, you get E
• If you need a value two positions to the right of E, you get B

This method is useful when your data represents something naturally cyclical, like time series with seasonal patterns or images that should tile seamlessly.

Symmetric = 1

Reflects the data at boundaries as if there were a mirror at each edge.

For Beginners: With Symmetric boundary handling, the data is reflected at the edges, as if there were mirrors placed at the boundaries.

Imagine standing between two mirrors and seeing your reflection repeated - the data is reflected back in the opposite direction when it hits a boundary.

Example: For a 1D array [A,B,C,D,E]:

• If you need a value one position to the left of A, you get B
• If you need a value two positions to the right of E, you get C

This method preserves continuity at the edges and is often used in image processing because it avoids creating artificial edges or discontinuities.

ZeroPadding = 2

Fills values outside the boundaries with zeros.

For Beginners: With ZeroPadding, any attempt to access data beyond the boundaries returns a zero value.

Imagine your data is surrounded by an infinite sea of zeros in all directions.

Example: For a 1D array [A,B,C,D,E]:

• If you need a value one position to the left of A, you get 0
• If you need a value one position to the right of E, you get 0

This method is simple to implement and understand. It's commonly used in signal processing and neural networks, especially in convolutional layers. However, it can create artificial edges at the boundaries where your data suddenly transitions to zeros.

Remarks

For Beginners: When working with operations like convolutions, filtering, or sampling, you often need to access data points outside the boundaries of your dataset.

For example, if you have an image and want to apply a filter to every pixel, what happens at the edges where the filter would need pixels that don't exist? Boundary handling methods provide different ways to solve this problem.

Think of it like trying to read a word at the edge of a page - you need to decide what to do when part of the word would be off the page.