Hybrid Dealiased Convolutions

Robert Joseph George,Noel Murasko,John C. Bowman

2023-05-14

Abstract:This paper proposes a practical and efficient solution for computing convolutions using hybrid dealiasing. It offers an alternative to explicit or implicit dealiasing and includes an optimized hyperparameter tuning algorithm that uses experience to find the optimal parameters. Machine learning algorithms and efficient heuristics are also developed to estimate optimal parameters for larger convolution problems using only small squares/rectangles.

Numerical Analysis

What problem does this paper attempt to address?

### What problem does this paper attempt to solve?

This paper aims to solve the problem of how to effectively avoid aliasing when calculating convolution. Specifically, the author proposes a new method - **Hybrid Dealiasing** - to replace explicit dealiasing and implicit dealiasing, and develops an optimized hyper - parameter adjustment algorithm to improve computational efficiency.

#### Background and Problem Description

1. **The relationship between convolution and the Fast Fourier Transform (FFT)**:
- The convolution theorem shows that the convolution of two functions can be efficiently calculated by the point - by - point product of their Fourier transforms.
- However, in practical applications, we usually need to calculate linear convolution rather than circular convolution. To ensure the correctness of linear convolution, the representation of the input signal in the frequency domain cannot overlap, otherwise aliasing errors will occur.

2. **Existing dealiasing methods**:
- **Explicit Dealiasing**: Avoid aliasing by padding zeros at the end of the input sequence. This method is simple but inefficient because it introduces a large number of unnecessary zero - value calculations.
- **Implicit Dealiasing**: Use the known zero - value positions to perform dealiasing without explicitly padding zeros. Although more efficient, it has strict requirements on the size of the input array and may not be applicable in all cases.

3. **Hybrid Dealiasing**:
- The author proposes a method that combines explicit and implicit dealiasing, called hybrid dealiasing. This method aims to take advantage of the speed advantage of implicit dealiasing while not being limited by the size of the input array.
- The key to hybrid dealiasing is how to efficiently determine the optimal padding length \( L_{\text{pad}} \) to ensure that computational resources are not wasted and aliasing can be effectively avoided.

4. **Hyper - parameter optimization**:
- To further improve computational efficiency, the author develops an empirical - based hyper - parameter adjustment algorithm. This algorithm estimates the optimal parameters through machine learning and heuristic methods, so that it can find the best solution more quickly when dealing with large - scale convolution problems.

#### Main Contributions of the Paper

- Proposes a new hybrid dealiasing method that can efficiently handle convolution problems in different dimensions (1D, 2D, 3D).
- Develops an optimized hyper - parameter adjustment algorithm that can significantly reduce computational time and memory usage.
- Verifies the effectiveness and superiority of the hybrid dealiasing method through experiments, especially in application scenarios such as image convolution.

In summary, this paper solves the problem of how to efficiently avoid aliasing when calculating convolution, and proposes the innovative method of hybrid dealiasing, providing new ideas and technical means for research in related fields.