Fraunhofer Institute for Computer Graphics Research IGD
Quantum Computing
Event details
Program
The Center of Applied Quantum Computing (ZAQC) offers a lecture series on selected topics in applied quantum computing in June. The three lectures will cover the topics
- "Quantum-inspired optimization" (Wednesday, June 4, 2025),
- "Tensor networks for physical simulation" (Thursday, June 12, 2025) and
- "Quantum computing for protein folding" (Wednesday, June 18, 2025).
The lectures will take place in a hybrid format in the auditorium of the Fraunhofer IGD, each at 5:00 PM. Please register for the event using the corresponding link and choose whether you would like to participate in person or virtually.
Location
Lecture hall of Fraunhofer IGD
Fraunhoferstr. 5
64283 Darmstadt
Date
June 04, 2025 - April 18, 2025, 5:00 PM
Language
English
Topics of the lectures
Quantum-inspired optimization
Quantum-inspired optimization
Quantum-inspired optimization leverages principles from quantum mechanics to enhance classical algorithms, offering novel approaches to solve complex binary optimization problems potentially more efficiently. Binary optimization has a wide range of applications including logistics, finance, machine learning or pharmacy.
In this talk we will give an overview over classical as well as quantum and quantum-inspired algorithms to heuristically solve binary optimization problems including simulated annealing, parallel tempering, quantum annealing and simulated quantum annealing. We will compare these algorithms and discuss their respective advantages and disadvantages.
Topics of the talk:
- Introduction to quadratic binary optimization
- Classical algorithms
- Quantum and quantum-inspired algorithms
Tensor networks for physical simulation
Tensor networks for physical simulation
The numerical simulation of physical properties is one of the most important challenges in science. If successful, it allows for the prediction of systems that have not been subject of experiments yet. However, the main problem of simulating physics is its ridiculous complexity. Most of the time, we have to deploy a multitude of approximations, and even then, a lot of calculations scale exponentially, either with the system size or with the accuracy needed for providing reasonable statements. In a lot of cases, the bad scaling of the simulations is owed to the so-called “Curse of dimensionality” which describes the exponential scaling of data in high-dimensional spaces. Solutions to this problem are well known and reach from sparse data structures to neural networks and, recently, quantum
computing. Since each solution has its own benefits and drawbacks, one has to carefully choose the right tool for the given problem.
One tool currently emerging is based on tensor networks. These have the capability of breaking the “Curse of dimensionality” by approximating an exponential amount of data as a specific set of low-dimensional tensors, which are connected to each other. They have been successfully used in a broad range of applications, ranging from quantum chemical calculations to engineering problems, and can be considered as a promising alternative to the aforementioned methods.
Topics of the talk:
- Introduction to tensor networks and their formats
- Quantics representation
- Functions as tensor networks
- Finite-Differences-Method
- Finite-Element-Method
Quantum computing for protein folding
Quantum computing for protein folding
The prediction of the three-dimensional structure of a given protein based solely on the information of its amino acid sequence is an important problem in drug discovery and computational biology. Even though recent breakthroughs in artificial intelligence-based models, such as DeepMind’s AlphaFold, have made a plethora of structures available, a significant fraction of proteins still elude these approaches. Especially proteins with limited training data or conformational changes during binding still provide significant hurdles.
In contrast, physics-based approaches, while not being limited to these problems, still lack behind. Part of the problem stems from the exponentially large space of possible configurations even for smaller peptides. Further, these configurations can be separated by steep energy barriers leading to many classical heuristics being trapped in local minima. Quantum optimization promises a novel solution to this problem, traversing steep energy barriers via the quantum tunnel effect. Still, current noisy intermediate scale Quantum computers (NISQ) are not sufficiently advanced to handle all-atom representations of the full protein. Are these computers able to predict the coarse-grained structure of proteins, even though they are currently limited? How do current-gen devices compare with well-established classical solvers?
Topics of the talk:
- Quantum optimization via quantum annealing
- Protein folding as an optimization problem
- Encoding of protein folding for NISQ hardware
- Energy landscapes and spin overlap distribution
- Comparison to classical solvers