PsiQuantum’s Post

Quantum Package 2.0: An Open-Source Determinant-Driven Suite of Programs https://lnkd.in/eyGR4xrT Abstract Quantum chemistry is a discipline which relies heavily on very expensive numerical computations. Therefore, performing accurate calculations on chemically meaningful systems requires (i) approximations that can lower the computational scaling and (ii) efficient implementations that take advantage of modern massively parallel architectures. Quantum Package is an open-source programming environment for quantum chemistry specially designed for wave function methods. Its main goal is the development of determinant-driven selected configuration interaction (sCI) methods and multireference second-order perturbation theory (PT2). The determinant-driven framework allows the programmer to include any arbitrary set of determinants in the reference space, hence providing greater methodological freedom. The sCI method implemented in Quantum Package is based on the CIPSI (Configuration Interaction using a Perturbative Selection made Iteratively) algorithm which complements the variational sCI energy with a PT2 correction. Additional external plugins have been recently added to perform calculations with multireference coupled cluster theory and range-separated density-functional theory. All the programs are developed with the IRPF90 code generator, which simplifies collaborative work and the development of new features. Quantum Package strives to allow easy implementation and experimentation of new methods, while making parallel computation as simple and efficient as possible on modern supercomputer architectures. Currently, the code enables, routinely, to realize runs on roughly 2 000 CPU cores, with tens of millions of determinants in the reference space. Moreover, we have been able to push up to 12 288 cores in order to test its parallel efficiency. In the present manuscript, we also introduce some key new developments: (i) a renormalized second-order perturbative correction for efficient extrapolation to the full CI limit and (ii) a stochastic version of the CIPSI selection performed simultaneously to the PT2 calculation at no extra cost.

This one impressed me. Cornell University researchers have described the quantum control machine, an instruction set architecture featuring conditional jump design, enabling developers to correctly express control flow via quantum algorithms, using a program counter in place of logic gates: https://lnkd.in/eSjdkZR4 This is novel, as typically one cannot correctly constitute control flow in quantum algorithms by utilizing classical conditional jump instruction in superposition; basically doing so defeats the quantum advantage. This "quantum control machine" approach is a potentially a large stride in advancement of its field.

Introduction: Studying single and parallel quantum algorithms at many different Epochs provides an indication of quantum benefit through the analysis of Loss, Accuracy, Runtime, and RAM results. In addition, more comprehensive algorithm comparisons were obtained through the analysis of Loss Efficiencies with a time component; along with the standard deviation of Loss Differences to measure volatility. First: A single quantum circuit was compared to two circuits in parallel across several experiments ranging from 1-15 Epochs. The main findings were that while the single circuit was favored in Runtime and RAM, the double circuit had best loss and accuracy. Most importantly, the larger circuit had better standard deviation of loss differences combined with higher training and validation loss efficiencies and 100% accuracies. (Slides 5-6) Second: This study featured 88 experiments with a fixed algorithm while varying the number of parallel quantum algorithms from 1-8. The findings were that 3 smaller parallel quantum architectures had top efficiencies across 1-50 epoch studies. The 7 layer circuit had the best standard deviation of loss differences, and the 8 layer circuit had the best overall loss values, likely associated with more processing time per epoch. The top rankings for layers 1-8 on slide 14 indicate a 20.0% increase in training loss efficiency (0.42/0.35), and a 14.9% increase in validation loss efficiency (0.54/0.47) from a single circuit to two circuits in parallel. Third: The circuit algorithms with 1) Best efficiency, 2) Lowest Runtime and RAM, and 3) Lowest loss from the second study were reconfigured to have similar run times using specific numbers of epochs. At over 200 epochs, the single circuit produced similar results to 2 parallel quantum architectures using the original model dataset. The rationale behind this was that parallel quantum architectures assisted the embedding with trainable parameter performance most optimally when runtimes were low to moderate; while the machine learning optimization process with more iterative dataset analysis was more pronounced at high runtimes. Six Installments of Parallel Quantum Algorithm Studies: See Attached PDF slide 15.

🚀 Qrisp 0.3 - Elevate Your #QuantumProgramming Experience! 🚀 We are thrilled to unveil the latest update to Qrisp, our cutting-edge quantum programming language. Qrisp 0.3 is here, packed with exciting features designed to supercharge your quantum programming tasks. Let's dive into the innovations that await you: 🔮 Quantum Approximate Optimization Algorithm (QAOA) Module 🔮 Our brand-new QAOA module seamlessly integrates this powerful quantum algorithm into Qrisp. Whether you're tackling industrial optimization problems or exploring quantum computing's potential, Qrisp's QAOA module has you covered. Highlights include: ✨ QAOAProblem Class: Simplify problem formulation and evaluation. ✨ QAOABenchmark Class: Analyze the performance of your implementations. ✨ Comprehensive Tutorials: From basics to advanced scientific concepts. ✨ 7 Presolved Optimization Problems: Jumpstart your optimization projects. ✨ Qiskit Runtime Integration: Utilize Qiskit runtime services as a Qrisp backend. 🌟 Plus, don't miss our novel mixer architecture, designed to maintain arbitrary problem constraints with polynomial complexity. Discover more about this exciting innovation here: ▶ https://lnkd.in/eVjW5rxZ 🔥 Upgraded Backtracking 🔥 Our backtracking module now boasts a 300% performance boost thanks to a new encoding method. With reduced circuit depth and fewer calls to the "reject" function, solving backtracking problems has never been faster or more efficient. 📚 Interactive Documentation 📚 Explore our fully interactive documentation, powered by the Thebe framework. We've also added some stylistic improvements for an enhanced learning experience. 🌐 Framework Interfacing 🌐 Qrisp Quantum Circuits can now be exported to Pennylane and PyTket, opening up new possibilities for quantum algorithm development and experimentation. Stay at the forefront of quantum computing with Qrisp 0.3! Upgrade today and unlock a world of quantum possibilities. Discover our improvements to Qrisp in detail here: ▶ https://lnkd.in/e6JPECwy #Qrisp #QuantumProgramming #QuantumComputing #Optimization #Industry #Innovation #Qrisp03 #Technology #Futurism Raphael Seidel Nikolay Tcholtchev, PhD Sebastian Bock Matic Petrič Niklas Steinmann Matteo Antonio Inajetovic Andreas Hoffmann Manfred Hauswirth Fraunhofer-Gesellschaft Bundesministerium für Wirtschaft und Klimaschutz

Solving Complex Problems with Evolutionary Computing: A Breakthrough in Problem-Solving Solving Complex Problems with Evolutionary Computing: A Breakthrough in Problem-Solving Introduction In today’s world, businesses and organizations face a multitude of complex problems that require innovative and efficient solutions. These problems often involve a large number of variables, constraints, and uncertainties, making them difficult to solve using traditional methods. However, with the advent of evolutionary […] https://lnkd.in/dATrqxix

🚀 DYK 20 - CAP Theorem: Consistency - A Tour of the BEAM (Elixir, Gleam, Erlang) on Fault-Tolerance, Concurrency, Parallelism, Distribution and Scalability Do you know what the letter C from the CAP Theorem means in distributed computing? The CAP Theorem says that in Consistency, it is expected that an operation is atomic or linearizable. This requires that operations in a distributed system be performed as if they were being done on a single node, by applying the operation across all nodes in a way that it looks like they all have been done simultaneously, not necessarily at the same exact nanosecond. This means that what one can expect in terms of Consistency is for operations to always output the same result. This requires that one operation, or more, that are working on the same data always have the same result in any node they run. Depending on the business case for the distributed system, this may require different mechanisms to guarantee data Consistency across it. For example, a distributed system that operates on data from an external database (reads and/or writes) may require different mechanisms to achieve Consistency compared with a distributed system that operates on data stored internally, on a per node basis, e.g., a local database. What one cannot expect is for operations to be instantaneous and global across a cluster of nodes, which is virtually impossible because the universe doesn't allow it in a distributed system over different physical locations, not even at the speed of light. Instead, delays can be expected in how an operation propagates into all nodes of a cluster and to see differences in performance as it is processed. Remember your database read replicas being behind the main node, the one that accepts the writes? This means that replications in a distributed database can only be considered strongly consistent when your application can afford some stale reads for a short period of time; otherwise, it needs to be treated as eventually consistent. Another example that affects Consistency is when you experience fragmented results in a distributed storage engine, one where data is partitioned across nodes. For example, returning different results for the same query because of a node timeout or being disconnected from the cluster. Achieving atomic operations in a distributed system, where data is replicated or spread across it, is not possible due to the physical limitations of the universe we live in. Therefore, the goal one may want to strive for in terms of Consistency is to find the balance where your application can work reliably, knowing that it cannot have the same atomic operations guarantees as a system running on a single node. The next DYK in this series will focus on letter A of the CAP Theorem. See you tomorrow 😀 📢 Like and repost to also reach your network. ✅ Follow me to see more like this on your feed 😀 👉 Leave your thoughts and questions below 👇 #BEAMDYK #Elixir #Erlang #MyelixirStatus

Generative Artificial Intelligence In the Loop for Very Large-Scale Cybernetics Software Systems Engineering using OpenAI Technologies. Design of Real-Time Schedulers for Very Large-Scale Cybernetics Networked Systems.

The calculation of the norm of vectors is essential in both artificial intelligence and quantum computing for tasks such as feature scaling, regularization, distance metrics, convergence criteria, representing quantum states, ensuring unitarity of operations, error correction, and designing quantum algorithms and circuits. DZone

As we all know, there has been much excitement about applications of quantum computing in finance, logistics, machine learning and AI. But we need to go beyond the hype and look at where we really are with quantum computing today and what are the applications that are within reach of near-term, pre-fault tolerant quantum hardware. Our co-founder and CTO/CSciO Toby Cubitt explores the hard realities and the real opportunities of quantum computing as it stands in this great piece in EE Times Europe. Learn more here: https://lnkd.in/eTsCh6X9 #quantumcomputing #hype #quantumalgorithims

Solving Complex Problems with Evolutionary Computing: A Breakthrough in Problem-Solving Solving Complex Problems with Evolutionary Computing: A Breakthrough in Problem-Solving Introduction In today’s world, businesses and organizations face a multitude of complex problems that require innovative and efficient solutions. These problems often involve a large number of variables, constraints, and uncertainties, making them difficult to solve using traditional methods. However, with the advent of evolutionary […] https://lnkd.in/dW_v3JfX