Abstract Floquet engineering provides a unique platform for realizing novel non-equilibrium topological phases. However, the intrinsic multi-gap structure of Floquet systems poses significant challenges for their classification using conventional analytical methods. Here, we propose an unsupervised machine learning framework that employs a kernel defined directly in momentum-time ( k , t ) space, constructed from Floquet–Bloch states. This data-driven approach does not require prior knowledge of topological invariants. By using a compact computational procedure that works directly with the system’s eigenstates, the method can identify topologically distinct phases associated with both the 0 -gap and the π -gap across various symmetry classes (1D AIII, 1D D, and 2D A). Our results demonstrate that key topological characteristics of Floquet systems are directly accessible from their eigenstates in ( k , t ) space, providing a tool for the systematic classification and study of non-equilibrium topological phases without a priori invariant information.
Abstract Floquet engineering provides a unique platform for realizing novel non-equilibrium topological phases. However, the intrinsic multi-gap structure of Floquet systems poses significant challenges for their classification using conventional analytical methods. Here, we propose an unsupervised machine learning framework that employs a kernel defined directly in momentum-time ( k , t ) space, constructed from Floquet–Bloch states. This data-driven approach does not require prior knowledge of topological invariants. By using a compact computational procedure that works directly with the system’s eigenstates, the method can identify topologically distinct phases associated with both the 0 -gap and the π -gap across various symmetry classes (1D AIII, 1D D, and 2D A). Our results demonstrate that key topological characteristics of Floquet systems are directly accessible from their eigenstates in ( k , t ) space, providing a tool for the systematic classification and study of non-equilibrium topological phases without a priori invariant information.
Floquet engineering offers an unparalleled platform for realizing novel non-equilibrium topological phases. However, the unique structure of Floquet systems, which includes multiple quasienergy gaps, poses a significant challenge to classification using conventional analytical methods. We propose a novel unsupervised machine learning framework that employs a kernel defined in momentum-time (k,t) space, constructed directly from Floquet-Bloch eigenstates. This approach is intrinsically data-driven and requires no prior knowledge of the underlying topological invariants, providing a fundamental advantage over prior methods that rely on abstract concepts like the micromotion operator or homotopic transformations. Crucially, this work successfully reveals the intrinsic topological characteristics encoded within the Floquet eigenstates themselves. We demonstrate that our method robustly and simultaneously identifies the topological invariants associated with both the 0-gap and the π-gap across various symmetry classes (1D AIII, 1D D, and 2D A), establishing a robust methodology for the systematic classification and discovery of complex non-equilibrium topological matter.
Abstract Recent experiments have successfully realized multi-band non-Abelian topological insulators with parity-time symmetry. Their topological classification transcends the conventional ten-fold classification, necessitating the use of non-Abelian groups, manifesting novel properties that cannot be described using integer topological invariants. The unique non-commutative multiplication of non-Abelian groups, along with the distinct topological classifications in the context of homotopy with or without a fixed base point, makes the identification of different non-Abelian topological phases more nuanced and challenging than in the Abelian case. In this work, we present an unsupervised learning method based on diffusion maps to classify non-Abelian multi-gap topological phases. The automatically adiabatic pathfinding process in our method can correctly sort the samples in the same phase even though they are not connected by adiabatic paths in the sample set. Most importantly, our method can deduce the multiplication table of the non-Abelian topological charges in a data-driven manner without requiring a priori knowledge. Additionally, our algorithm can provide the correct classifications for the samples within both the homotopy with and without a fixed base point. Our results provide insights for future studies on non-Abelian phases using machine learning approaches.
A remarkable breakthrough in topological phase classification is the establishment of the topological periodic table, which is mainly based on the classifying space analysis or K theory, but not based on concrete Hamiltonians that possess finite bands or arise in a lattice. As a result, it is still difficult to identify the topological phase of an arbitrary Hamiltonian; the common practice is, instead, to check the incomplete and still growing list of topological invariants one by one, very often by trial and error. Here, we develop unsupervised classifications of topological gapped systems with symmetries, and demonstrate the data-driven construction of the topological periodic table without a priori knowledge of topological invariants. This unsupervised data-driven strategy can take into account spatial symmetries, and further classify phases that were previously classified as trivial in the past. Our Letter introduces machine learning into topological phase classification and paves the way for intelligent explorations of new phases of topological matter.
The discovery of topological phases in non-equilibrium quantum systems, particularly those driven by periodic fields (Floquet systems), holds immense potential for next-generation quantum technologies. However, identifying and classifying these phases is notoriously challenging, as the associated topological invariants are often non-local, abstract, and lack straightforward experimental observables. Traditional theoretical methods typically rely on idealized models and exact symmetries, which break down in the presence of disorder, interactions, or finite-size effects inherent to real materials. Here, we propose a novel paradigm for the discovery of hidden Floquet topological invariants using hybrid quantum-classical neural networks. Our framework synergistically combines a parameterized quantum circuit, acting as a variational ansatz for the Floquet quasienergy states, with a classical neural network that learns to map the system's time-resolved dynamics onto a robust, data-driven topological classifier. By training the hybrid model on simulated measurement data—such as time-of-flight images or spin-resolved photoelectron spectra—the classical network distills the essential, often obscured, features of the system's unitary evolution without prior knowledge of the underlying Hamiltonian. Crucially, we demonstrate that this approach can uncover novel invariants for anomalous Floquet topological phases that are invisible to conventional frequency-domain analyses. Furthermore, the trained network provides a practical, noise-resilient protocol for real-time classification, enabling the experimental detection of topological transitions solely from short-time dynamics. This work establishes a powerful, generalizable framework for accelerating the discovery and characterization of emergent topological order in complex driven quantum matter, bridging the gap between abstract topological theory and experimental reality.
The integration of machine learning into fundamental science has opened new avenues for addressing long-standing challenges rooted in mathematical limitations. For instance, while topological invariants are essential for characterizing topological phases of matter, no single invariant is universally applicable. This limitation explains why, over decades of classifying topological phases-primarily in Hermitian systems-many phases initially deemed 'trivial' were later recognized as topological. Recently, the discovery of non-Hermitian band topology has driven substantial efforts in non-Hermitian topological classification, leading to the development of new topological invariants. However, these invariants still fail to capture all non-Hermitian topological features. Here, without relying on any topological invariant, we develop a machine-learning algorithm for the unsupervised classification of symmetry-protected non-Hermitian topological phases. By utilizing random Hamiltonians, we unsupervisedly construct a topological periodic table without requiring advanced mathematical knowledge. Furthermore, based on the learning results, we derive a formula that reveals the impact of parity transformation on periodicity. Our algorithm can also account for boundary effects, enabling the exploration of open-boundary influences on the topological phase diagram. These findings establish an unsupervised approach for classifying symmetry-protected non-Hermitian topological phases, uncover previously unnoticed topological features in non-Hermitian systems, and provide valuable guidance for both theoretical advancements and experimental realizations.
Nonequilibrium topological matter has been a fruitful topic of both theoretical and experimental interest. A great variety of exotic topological phases unavailable in static systems may emerge under nonequilibrium situations. How to locate the borders between different nonequilibrium topological phases is an important issue to facilitate topological characterization and further understand phase transition behaviors. In this paper, we develop an unsupervised machine-learning protocol to distinguish between different Floquet (periodically driven) topological phases possessed by theoretical models, by incorporating the system's dynamics within one driving period, adiabatic deformation in the time dimension, plus the system's symmetry all into our machine-learning algorithm. Results from two rich case studies indicate that machine learning is able to reliably reveal intricate topological phase boundaries and can hence be a powerful tool to analyze and discover previously unknown topological phases afforded by the time dimension. As inspired by previous elegant methods in analyzing Floquet topological phases, one essential ingredient in our machine-learning approach is to directly investigate the system's evolution operators parametrized by time.