Angular Opposition Interpolation (AOI)
A geometric interpolation concept based on angular balance around a target point
Author: Joseph Weaver
Status: Early concept / research note
Priority: Back burner
1. Purpose
This document captures an interpolation concept so it is not forgotten while other work takes priority.
The idea is motivated by a simple geometric observation:
Interpolation is most stable when the target point is surrounded by observations.
Many standard interpolation methods (for example inverse distance weighting) only consider distance, but not whether surrounding points are balanced in direction.
The goal of this concept is to construct an interpolation scheme that rewards point configurations that surround the target point and penalizes configurations where all points lie on one side.
The key idea is to weight observations according to
- distance from the target
- angular separation relative to other points
In particular, points that lie opposite each other across the target location should reinforce each other.
2. Problem Setup
Suppose we have observations
(x_i, y_i, z_i), i = 1,…,n
and we want to estimate
z(x,y)
at location
s₀ = (x,y).
Define polar coordinates relative to the target.
Distance
d_i = sqrt((x_i - x)^2 + (y_i - y)^2)
Angle
a_i = atan2(y_i - y, x_i - x)
Each observation therefore has a distance d_i and direction a_i relative to the target point.
3. Geometric Intuition
Consider two points.
If the two points lie
• in the same direction from the target, they provide redundant information
• in opposite directions, they provide balanced information
Example geometry:
Case 1 (bad geometry)
P1 —-> Target
P2 —->
Both observations lie on the same side.
Case 2 (good geometry)
P1 —-> Target <—- P2
The target is bracketed by the observations.
Therefore interpolation should reward angular opposition.
4. Angular Separation
Define angular separation between two points
| Δθ_ij = min( | a_i − a_j | , 2π − | a_i − a_j | ) |
This ensures
0 ≤ Δθ_ij ≤ π
Examples
same direction → 0
perpendicular → π/2
opposite → π
Normalize this measure
A_ij = Δθ_ij / π
Then
same direction → 0
perpendicular → 0.5
opposite → 1
This value represents degree of opposition.
5. Two Point Interpolation Concept
Suppose we begin with the nearest observation
(x₁, y₁)
with weight
w₁ = 1
Now introduce a second observation
(x₂, y₂)
with distance d₂ and angle a₂.
Compute angular separation
| Δθ = min( | a₂ − a₁ | , 2π − | a₂ − a₁ | ) |
Define opposition factor
A = Δθ / π
Assign weight to the second point
w₂ = (d₁ / (d₁ + d₂)) × A
Update
w₁ = 1 − w₂
Interpretation
If point 2 lies in the same direction as point 1, A = 0, so it receives no weight.
If the points are opposite, A = 1 and weights depend only on relative distances.
6. Sequential Weight Redistribution
One possible algorithm is sequential.
- Find nearest neighbor
- Assign weight 1
- Add neighbors sequentially
- Redistribute weights using angular-distance rule
Pseudo algorithm
sort points by distance from target
set w₁ = 1
for each new point i
compute angular separation with existing points
compute angular opposition score
transfer weight according to distance and angle
normalize weights
This produces a competitive redistribution process where new observations gain influence only if they improve directional balance.
7. Alternative Global Formulation
A more stable formulation may compute weights simultaneously.
Distance kernel
D_i = 1 / (d_i + ε)^p
Angular contribution
A_i = sum over j ≠ i of (Δθ_ij / π) × D_j
Raw weight
w_i* = D_i × A_i
Normalize
w_i = w_i* / sum(w_k*)
Interpolation
ẑ(x,y) = sum( w_i × z_i )
8. Relationship to Existing Methods
This concept relates to several known interpolation methods.
Inverse Distance Weighting (IDW)
Weights depend only on distance.
Weakness: points on one side may dominate.
Sector or Quadrant Interpolation
Used in geostatistics.
Space is divided into angular sectors and neighbors are selected from each sector.
Natural Neighbor Interpolation
Based on Voronoi diagrams.
Weights depend on how much Voronoi area the target steals from neighbors.
Kriging
Weights derived from spatial covariance models.
9. Potential Advantages
Possible benefits
• Encourages directional balance
• Reduces bias when points lie on one side
• No covariance model required
• Simple geometric interpretation
• Works with irregular point distributions
10. Potential Weaknesses
Possible limitations
• Requires pairwise angular calculations
• May become unstable if observations cluster heavily
• Sequential updates could depend on ordering
• Requires empirical validation
11. Possible Names
Working names for the method
Angular Opposition Interpolation (AOI)
Angularly Balanced IDW
Opposition Weighted Interpolation
Directional Shepard Interpolation
12. Possible Experiments
Compare AOI against
• Inverse Distance Weighting
• Natural Neighbor Interpolation
• Kriging
• Radial Basis Function interpolation
Evaluation metrics
• RMSE
• cross validation error
• robustness to irregular sampling
Test surfaces
• smooth surfaces
• ridges or gradients
• clustered observations
• sparse sampling
13. Future Work
Possible theoretical directions
- Connect to harmonic interpolation
- Interpret as polar kernel smoothing
- Analyze bias and variance
- Explore relationship to discrete Laplacian operators
14. Summary
The central idea is simple.
Interpolation should reward point configurations that surround the target location.
This concept implements that idea by weighting observations using
distance from the target and angular opposition relative to other observations.
The method is related to existing interpolation approaches but introduces explicit geometric weighting based on opposing directions.
Further testing is required to determine whether the method improves predictive performance relative to established approaches.
End of note.
Comments