🔢 mathematics

t-Product and t-STP of Cubic Matrices With Application to Hyper-Networked Systems

This paper introduces the t-semi-tensor product (t-STP) for cubic matrices to overcome the dimensional and coupling limitations of existing algebraic operations, establishing a comprehensive algebraic framework that enables the modeling of complex, coupled dynamic control systems in hyper-networked applications such as supply chain evolutionary games.

Original authors: Daizhan Cheng, Zhengping Ji

Published 2026-04-08

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Daizhan Cheng, Zhengping Ji

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to organize a massive, multi-layered library.

In the old days, we only had flat books (2D matrices). You could stack them, read them, and mix their pages easily. But the real world isn't flat. It's 3D. Think of a Rubik's Cube or a stack of transparent sheets where each sheet has a different picture, but the pictures are related. This is what mathematicians call a Cubic Matrix (or a 3D block of data).

The problem is: How do you do math with these 3D blocks?

This paper is like a new instruction manual for "3D Math." It introduces three new ways to multiply these blocks, solving a puzzle that has been bothering scientists for years.

Here is the story of the three "tools" they invented, explained simply:

1. The Strict Rule (The t-Product)

Imagine you have a stack of 100 sheets of paper. The t-Product is a very strict librarian.

The Rule: "I will only let you mix Sheet A with Sheet B if they are exactly the same size and perfectly aligned."
The Good: It's very organized. It treats the whole stack as one giant, connected unit. If you change the top sheet, the bottom sheet knows about it.
The Bad: It's too picky. If you have a small sheet and a big sheet, the librarian says, "Nope, I can't do that." This limits what you can model in the real world, where things are often different sizes.

2. The Flexible Rule (The DK-STP)

To fix the "too picky" problem, the authors invented a second tool called DK-STP (Dimension-Keeping Semi-Tensor Product).

The Rule: "I don't care about the size! I can mix a tiny sheet with a huge sheet."
The Good: It's incredibly flexible. It can handle messy, real-world data of any shape.
The Bad: It's too independent. Imagine you have 100 sheets. This tool treats every single sheet as if it's in a different room, with no doors between them. If you change the top sheet, the bottom sheet has no idea it happened. In the real world (like a supply chain), if a factory changes its output, the market should know. This tool breaks that connection.

3. The Perfect Hybrid (The t-STP)

This is the paper's main invention. The authors realized they needed a tool that had the flexibility of the second tool but the connection of the first.

The Solution: They created the t-STP (t-Semi-Tensor Product).
How it works: It's like a "Smart Librarian." It says, "I can mix sheets of different sizes (like the flexible one), BUT I will also build a hallway between them so they can talk to each other (like the strict one)."
Why it matters: Now, you can model complex systems where different parts are different sizes, but they still influence each other.

The Real-World Example: The Supply Chain Game

To prove this new math works, the authors applied it to a Supply Chain (think: Factories $\to$ Wholesalers $\to$ Markets).

The Old Way: You might try to model this as a flat list. But a supply chain is a web. A factory affects multiple wholesalers, who affect multiple markets. It's a 3D relationship.
The New Way: They used the t-STP to build a "Hyper-Network."
- Imagine the Factories are the first layer of the cube.
- The Wholesalers are the second layer.
- The Markets are the third layer.
- The t-STP allows them to calculate how a change in one Factory ripples through the Wholesalers and finally changes the Markets, even if the numbers of factories and markets are different.

They treated this whole system like a game (an "Evolutionary Game"). Just like in a video game where your move changes the whole board, their math shows how one company's decision changes the entire global supply network.

Why Should You Care?

Handling Big Data: We live in an era of massive data (AI, 3D imaging, complex networks). This math gives us a way to crunch that data without getting bogged down by rigid rules.
Better Control: If you are designing a self-driving car or a power grid, you need to understand how different parts interact. This new math allows engineers to build better control systems that are more robust and flexible.
New Algebra: The authors didn't just make a calculator; they built a whole new "language" (Groups, Rings, Lie Algebras) for 3D data. This is like inventing a new grammar for a language that didn't exist before.

The Bottom Line

This paper is about breaking down the walls between different types of data.

Before, you had to force 3D data into 2D boxes (losing information).
Or, you had to treat 3D data as isolated islands (losing connection).
Now, with the t-STP, we have a tool that lets 3D data be flexible in size but connected in action. It's a new superpower for understanding our complex, multi-layered world.

1. Problem Statement

The paper addresses limitations in modeling dynamic control systems using cubic matrices (order-3 hypermatrices), which are essential for representing higher-order data structures in fields like signal processing, computer vision, and complex networked systems.

Two existing algebraic frameworks for cubic matrices have significant drawbacks:

The t-Product: Proposed by Kilmer et al., this operation is powerful for tensor analysis but imposes strict dimensional constraints. It requires the inner dimensions of the cubic matrices to match exactly (e.g., $A \in \mathbb{R}^{m \times n \times s}$ and $B \in \mathbb{R}^{n \times p \times s}$ ), limiting modeling flexibility for systems with varying dimensions.
The Dimension-Keeping Semi-Tensor Product (DK-STP): An extension of the classical Semi-Tensor Product (STP) that allows multiplication of matrices with arbitrary dimensions without expanding the dimension. However, when applied to cubic matrices, the DK-STP results in decoupled subsystem dynamics. Specifically, it treats frontal slices independently, failing to capture the interactions between different slices (subsystems) that are crucial for modeling complex, coupled systems.

Core Challenge: There is a need for a unified algebraic operation that allows for arbitrary dimensions (like DK-STP) while preserving coupling between frontal slices (like the t-product) to model complex, nonlinear, and hyper-networked dynamic systems.

2. Methodology

The authors propose a novel algebraic framework built on three main pillars:

A. Review and Extension of Existing Products

t-Product: Defined via the block circulant matrix ($bcirc$) and the unfold operation. The paper formally proves the associativity of the t-product, establishing that cubic matrices under this product form a ring, a module, and a monoid.
DK-STP: Defined by applying the standard DK-STP to each frontal slice independently. The paper demonstrates that while this removes dimensional constraints, it leads to a system where the $k$ -th frontal slice evolves independently of others ( $\dot{x}^{(k)} = A^{(k)} \star x^{(k)}$ ), which is insufficient for coupled systems.

B. Proposal of the t-Semi-Tensor Product (t-STP)

To resolve the trade-off between dimension flexibility and subsystem coupling, the authors introduce the t-STP ( $\ltimes^*$ ).

Definition: The t-STP integrates the t-product and DK-STP. For two cubic matrices $A$ and $B$ , the operation is defined as:
$A \ltimes^* B := \Gamma(A) \ltimes B$
where $\Gamma(A) = bcirc(A)$ is the universal homomorphism mapping the cubic matrix to a block-circulant matrix, and $\ltimes$ is the DK-STP.
Mechanism: By first expanding the matrix into its block-circulant form (via $\Gamma$ ), the operation captures the cyclic interactions between slices (coupling). Then, the DK-STP is applied to handle arbitrary dimensions.
Properties: The paper proves that the set of all cubic matrices under t-STP forms a semi-group. Furthermore, it establishes that t-STP is a generalization of both the t-product (when dimensions match) and the DK-STP (when coupling is not required).

C. Algebraic Structure Construction

The authors systematically derive the algebraic structures over cubic matrices using t-STP:

Ring and Module: $(\mathbb{R}^{m \times n \times s}, +, \ltimes^*)$ forms a ring, and cubic matrices form a left module over this ring.
Lie Algebra and Lie Group:
- A Lie bracket is defined as $[A, B]_* = A \ltimes^* B - B \ltimes^* A$ .
- This forms a General Linear Lie Algebra, denoted $gl^*(m \times n \times s, \mathbb{R})$ .
- The corresponding invertible elements form a General Linear Lie Group, $GL^*(m \times n \times s, \mathbb{R})$ .
Universal Homomorphism: The map $\Gamma$ is shown to be a homomorphism from the t-STP structure to the standard matrix product structure, allowing complex tensor problems to be mapped to standard linear algebra problems for analysis.

D. Dynamic System Formulation

The authors construct t-STP based dynamic (control) systems:

Linear Systems: $\dot{x}(t) = A \ltimes^* x(t) + B \ltimes u(t)$ .
Nonlinear Systems: $\dot{x}(t) = f_*(x(t)) + \sum g_i(x(t)) u_i(t)$ .
Solution: The trajectory is solved using the t-function (e.g., $exp_*$ ), defined via Taylor series expansion using the t-STP power. The solution is shown to be bi-similar to standard Euclidean control systems via the $\Gamma$ map.

3. Key Contributions

Novel Algebraic Operation (t-STP): The introduction of the t-Semi-Tensor Product, which successfully unifies the dimension-flexibility of DK-STP with the slice-coupling capability of the t-product.
Algebraic Theory: The rigorous derivation of algebraic structures (groups, rings, modules, Lie algebras, and Lie groups) for cubic matrices under the t-STP, providing a theoretical foundation for higher-order control theory.
Dynamic System Framework: The construction of linear and nonlinear dynamic control systems over cubic matrices that allow for arbitrary dimensions and coupled subsystems.
Application to Hyper-Networks: The successful application of this framework to model hyper-networked evolutionary games, specifically a multi-tiered supply chain involving manufacturers, wholesalers, and markets.

4. Results

Associativity Proof: The paper provides a rigorous proof that the t-product is associative, a property often assumed but not explicitly proven in prior literature.
Coupling Verification: Through the t-STP formulation, the resulting dynamic equations show that the evolution of one frontal slice depends on all other slices, unlike the DK-STP where slices are independent.
Model Reduction: The framework demonstrates significant potential for model dimension reduction. In classical linear systems, a state $X \in \mathbb{R}^{m \times n \times s}$ would require a transition matrix of size $mns \times mns$ . Under the t-STP framework, the transition matrix is represented by a cubic matrix of dimension $ms \times p$ (or similar), significantly reducing computational complexity for large-scale data.
Supply Chain Simulation: The application to a supply network (2 manufacturers, 3 wholesalers, 4 markets) successfully models the interactions as a hyper-graph. The system is formulated as a dynamic control system where strategies evolve based on payoff gradients, demonstrating the practical utility of the t-STP in complex networked games.

5. Significance

Theoretical Advancement: This work bridges the gap between tensor algebra (t-product) and semi-tensor product theory (STP), creating a unified language for higher-order data dynamics. It extends the concept of Lie groups and algebras to non-square, higher-order tensors.
Practical Utility: The framework is particularly valuable for Big Data and Hyper-Networked Systems where data is naturally multi-dimensional and interactions are complex. The ability to handle arbitrary dimensions without exploding the state space makes it ideal for modern control problems in IoT, supply chains, and biological networks.
Future Directions: The paper suggests that these techniques can be extended to hypermatrices of order $d > 3$ and applied to nonlinear geometric control, offering a new pathway for analyzing high-dimensional dynamic systems.

In summary, the paper provides a robust mathematical toolkit for controlling and analyzing complex systems represented by cubic matrices, overcoming previous limitations regarding dimension constraints and subsystem decoupling.