On Complexity of Stability Analysis in Higher-order Ecological Networks through Tensor Decompositions

Anqi Dong,Can Chen

2024-04-03

Abstract:Complex ecological networks are often characterized by intricate interactions that extend beyond pairwise relationships. Understanding the stability of higher-order ecological networks is salient for species coexistence, biodiversity, and community persistence. In this article, we present complexity analyses for determining the linear stability of higher-order ecological networks through tensor decompositions. We are interested in the higher-order generalized Lotka-Volterra model, which captures high-order interactions using tensors of varying orders. To efficiently compute Jacobian matrices and thus determine stability in large ecological networks, we exploit various tensor decompositions, including higher-order singular value decomposition, Canonical Polyadic decomposition, and tensor train decomposition, accompanied by in-depth computational and memory complexity analyses. We demonstrate the effectiveness of our framework with numerical examples.

Systems and Control,Numerical Analysis,Optimization and Control

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the problem of determining linear stability through tensor decomposition in high - order ecological networks. Specifically, the researchers focus on how to efficiently calculate the Jacobian matrix in complex ecological networks containing high - order interactions and further analyze the stability of these systems. The paper mentions that traditional ecological models usually only consider pairwise interactions between species and ignore higher - order interactions, and these higher - order interactions may have an important impact on species coexistence, biodiversity and community persistence in ecosystems. In order to better understand and analyze these high - order interactions, the researchers introduced the high - order generalized Lotka - Volterra (HOGLV) model and used different tensor decomposition methods (such as high - order singular value decomposition HOSVD, Canonical Polyadic decomposition CPD and tensor train decomposition TTD) to improve the efficiency of calculating the Jacobian matrix, thereby effectively evaluating the linear stability of the system.

### Core problems of the paper
1. **Stability analysis of ecological networks with high - order interactions**: How to analyze the stability of ecological networks when considering high - order interactions.
2. **Efficient calculation of the Jacobian matrix**: How to use tensor decomposition techniques (HOSVD, CPD, TTD) to efficiently calculate the Jacobian matrix to reduce computational complexity and memory requirements.

### Main methods
- **High - order generalized Lotka - Volterra (HOGLV) model**: Used to describe the dynamics of ecological networks containing high - order interactions.
- **Tensor decomposition**:
- **High - order singular value decomposition (HOSVD)**: Decomposes a high - order tensor into a core tensor and multiple factor matrices to reduce computational complexity.
- **Canonical Polyadic decomposition (CPD)**: Decomposes a high - order tensor into the sum of multiple rank - one tensors, which is suitable for low - rank approximation.
- **Tensor train decomposition (TTD)**: Decomposes a high - order tensor into a series of third - order core tensors, which has numerical stability and quasi - optimal low - rank approximation.

### Key formulas
- **Dynamics equation of the HOGLV model**:
\[
\dot{x}_i = x_i \left( r_i + \sum_{j = 1}^n A_{ij} x_j + \sum_{j = 1}^n \sum_{k = 1}^n B_{ijk} x_j x_k + \sum_{j = 1}^n \sum_{k = 1}^n \sum_{l = 1}^n C_{ijkl} x_j x_k x_l+\cdots \right)
\]
where \( r_i \) is the intrinsic growth rate of species \( i \), \( A \) is the second - order interaction matrix, and \( B \) and \( C \) are the third - order and fourth - order interaction tensors respectively.

- **Calculation of the Jacobian matrix**:
\[
M = X^* \left( A + B\times_2 x^*+ B\times_3 x^*+ C\times_{23} \{x^*, x^*\}+ C\times_{24} \{x^*, x^*\}+ C\times_{34} \{x^*, x^*\}+\cdots \right)
\]
where \( X^* \) is a diagonal matrix whose diagonal elements are \( x^*\).

### Contributions of the paper
- **Theoretical analysis**: Analyzed in detail the computational complexity and memory complexity of different tensor decomposition methods when calculating the Jacobian matrix.
- **Numerical experiments**: Verified the effectiveness of the proposed framework through specific numerical examples and demonstrated the advantages of different tensor decomposition methods in dealing with large - scale ecological networks.

Through these methods and analyses, the paper provides new tools and perspectives for understanding the stability of high - order ecological networks and helps to promote research in related fields such as ecology and systems biology.