arXiv subjects

Hristo Djidjev

Publications and source records attributed to Hristo Djidjev.

18 recordsLinked to original sources

arXiv2024

Increasing the Hardness of Posiform Planting Using Random QUBOs for Programmable Quantum Annealer Benchmarking

Posiform planting is a method for constructing QUBO problems with a single unique planted solution that can be tailored to arbitrary connectivity graphs. In this study we investigate making posiform planted QUBOs computationally harder by fusing many smaller random discrete coefficient spin-glass Ising models, whose global minimum energy is computed classically using classical binary integer programming optimization software, with posiform-planted QUBOs. The single unique ground-state solution of the resulting QUBO problem is the concatenation of (exactly one of) the ground-states of each of the smaller problems. We apply these modified posiform planted QUBOs to the task of benchmarking programmable D-Wave quantum annealers. The proposed method enables generating binary variable combinatorial optimization problems that cover the entire quantum annealing processor hardware graph, have a unique solution, are entirely hardware-graph-native, and can have tunable computational hardness. We benchmark the capabilities of three D-Wave superconducting qubit quantum annealing processors, having from 563 up to 5627 qubits, to sample the optimal unique planted solution of problems generated by our proposed method and compare them against simulated annealing and Gurobi. We find that the D-Wave QPU ground-state sampling success rate does not change with respect to the size of the random QUBOs we employ. Surprisingly, we find that some of these classes of QUBOs are solved at very high success rates at short annealing times compared to longer annealing times for the Zephyr connectivity graph QPUs.

quant-ph↗

arXiv2023

Initial State Encoding via Reverse Quantum Annealing and h-gain Features

Quantum annealing is a specialized type of quantum computation that aims to use quantum fluctuations in order to obtain global minimum solutions of combinatorial optimization problems. Programmable D-Wave quantum annealers are available as cloud computing resources which allow users low level access to quantum annealing control features. In this paper, we are interested in improving the quality of the solutions returned by a quantum annealer by encoding an initial state into the annealing process. We explore two D-Wave features allowing one to encode such an initial state: the reverse annealing and the h-gain features. Reverse annealing (RA) aims to refine a known solution following an anneal path starting with a classical state representing a good solution, going backwards to a point where a transverse field is present, and then finishing the annealing process with a forward anneal. The h-gain (HG) feature allows one to put a time-dependent weighting scheme on linear ($h$) biases of the Hamiltonian, and we demonstrate that this feature likewise can be used to bias the annealing to start from an initial state. We also consider a hybrid method consisting of a backward phase resembling RA, and a forward phase using the HG initial state encoding. Importantly, we investigate the idea of iteratively applying RA and HG to a problem, with the goal of monotonically improving on an initial state that is not optimal. The HG encoding technique is evaluated on a variety of input problems including the edge-weighted Maximum Cut problem and the vertex-weighted Maximum Clique problem, demonstrating that the HG technique is a viable alternative to RA for some problems. We also investigate how the iterative procedures perform for both RA and HG initial state encoding on random whole-chip spin glasses with the native hardware connectivity of the D-Wave Chimera and Pegasus chips.

quant-ph↗

arXiv2022

Advanced unembedding techniques for quantum annealers

The D-Wave quantum annealers make it possible to obtain high quality solutions of NP-hard problems by mapping a problem in a QUBO (quadratic unconstrained binary optimization) or Ising form to the physical qubit connectivity structure on the D-Wave chip. However, the latter is restricted in that only a fraction of all pairwise couplers between physical qubits exists. Modeling the connectivity structure of a given problem instance thus necessitates the computation of a minor embedding of the variables in the problem specification onto the logical qubits, which consist of several physical qubits "chained" together to act as a logical one. After annealing, it is however not guaranteed that all chained qubits get the same value (-1 or +1 for an Ising model, and 0 or 1 for a QUBO), and several approaches exist to assign a final value to each logical qubit (a process called "unembedding"). In this work, we present tailored unembedding techniques for four important NP-hard problems: the Maximum Clique, Maximum Cut, Minimum Vertex Cover, and Graph Partitioning problems. Our techniques are simple and yet make use of structural properties of the problem being solved. Using Erdős-Rényi random graphs as inputs, we compare our unembedding techniques to three popular ones (majority vote, random weighting, and minimize energy). We demonstrate that our proposed algorithms outperform the currently available ones in that they yield solutions of better quality, while being computationally equally efficient.

quant-ph↗

arXiv2022

Distributed Out-of-Memory SVD on CPU/GPU Architectures

We propose an efficient, distributed, out-of-memory implementation of the truncated singular value decomposition (t-SVD) for heterogeneous (CPU+GPU) high performance computing (HPC) systems. Various implementations of SVD have been proposed, but most only estimate the singular values as an estimation of the singular vectors which can significantly increase the time and memory complexity of the algorithm. In this work, we propose an implementation of SVD based on the power method, which is a truncated singular values and singular vectors estimation method. Memory utilization bottlenecks seen in the power method are typically associated with the computation of the Gram matrix $\mat{A}^T\mat{A}$, which can be significant when $\mat{A}$ is large and dense, or when $\mat{A}$ is super-large and sparse. The proposed implementation is optimized for out-of-memory problems where the memory required to factorize a given matrix is greater than the available GPU memory. We reduce the memory complexity of $\mat{A}^T\mat{A}$ by using a batching strategy where the intermediate factors are computed block by block. We also suppress I/O latency associated with both host-to-device (H2D) and device-to-host (D2H) batch copies by overlapping each batch copy with compute using CUDA streams. Furthermore, we use optimized \textit{NCCL} based communicators to reduce the latency associated with collective communications (both intra-node and inter-node). In addition, sparse and dense matrix multiplications are significantly accelerated with GPU cores (or tensors cores when available), resulting in an implementation with good scaling. We demonstrate the scalability of our distributed out of core SVD algorithm to successfully decompose dense matrix of size 1TB and sparse matrix of size 128PB with 1e-6 sparsity.

cs.DC↗

arXiv2022

Quantum Algorithm Implementations for Beginners

As quantum computers become available to the general public, the need has arisen to train a cohort of quantum programmers, many of whom have been developing classical computer programs for most of their careers. While currently available quantum computers have less than 100 qubits, quantum computing hardware is widely expected to grow in terms of qubit count, quality, and connectivity. This review aims to explain the principles of quantum programming, which are quite different from classical programming, with straightforward algebra that makes understanding of the underlying fascinating quantum mechanical principles optional. We give an introduction to quantum computing algorithms and their implementation on real quantum hardware. We survey 20 different quantum algorithms, attempting to describe each in a succinct and self-contained fashion. We show how these algorithms can be implemented on IBM's quantum computer, and in each case, we discuss the results of the implementation with respect to differences between the simulator and the actual hardware runs. This article introduces computer scientists, physicists, and engineers to quantum algorithms and provides a blueprint for their implementations.

cs.ET↗

arXiv2022

Distributed non-negative RESCAL with Automatic Model Selection for Exascale Data

With the boom in the development of computer hardware and software, social media, IoT platforms, and communications, there has been an exponential growth in the volume of data produced around the world. Among these data, relational datasets are growing in popularity as they provide unique insights regarding the evolution of communities and their interactions. Relational datasets are naturally non-negative, sparse, and extra-large. Relational data usually contain triples, (subject, relation, object), and are represented as graphs/multigraphs, called knowledge graphs, which need to be embedded into a low-dimensional dense vector space. Among various embedding models, RESCAL allows learning of relational data to extract the posterior distributions over the latent variables and to make predictions of missing relations. However, RESCAL is computationally demanding and requires a fast and distributed implementation to analyze extra-large real-world datasets. Here we introduce a distributed non-negative RESCAL algorithm for heterogeneous CPU/GPU architectures with automatic selection of the number of latent communities (model selection), called pyDRESCALk. We demonstrate the correctness of pyDRESCALk with real-world and large synthetic tensors, and the efficacy showing near-linear scaling that concurs with the theoretical complexities. Finally, pyDRESCALk determines the number of latent communities in an 11-terabyte dense and 9-exabyte sparse synthetic tensor.

cs.DC↗

arXiv2020

Inferring the Dynamics of the State Evolution During Quantum Annealing

To solve an optimization problem using a commercial quantum annealer, one has to represent the problem of interest as an Ising or a quadratic unconstrained binary optimization (QUBO) problem and submit its coefficients to the annealer, which then returns a user-specified number of low-energy solutions. It would be useful to know what happens in the quantum processor during the anneal process so that one could design better algorithms or suggest improvements to the hardware. However, existing quantum annealers are not able to directly extract such information from the processor. Hence, in this work we propose to use advanced features of D-Wave 2000Q to indirectly infer information about the dynamics of the state evolution during the anneal process. Specifically, D-Wave 2000Q allows the user to customize the anneal schedule, that is, the schedule with which the anneal fraction is changed from the start to the end of the anneal. Using this feature, we design a set of modified anneal schedules whose outputs can be used to generate information about the states of the system at user-defined time points during a standard anneal. With this process, called "slicing", we obtain approximate distributions of lowest-energy anneal solutions as the anneal time evolves. We use our technique to obtain a variety of insights into the annealer, such as the state evolution during annealing, when individual bits in an evolving solution flip during the anneal process and when they stabilize, and we introduce a technique to estimate the freeze-out point of both the system as well as of individual qubits.

quant-ph↗

arXiv2020

Advanced anneal paths for improved quantum annealing

Advances in quantum annealing technology make it possible to obtain high quality approximate solutions of important NP-hard problems. With the newer generations of the D-Wave annealer, more advanced features are available which allow the user to have greater control of the anneal process. In this contribution, we study how such features can help in improving the quality of the solutions returned by the annealer. Specifically, we focus on two of these features: reverse annealing and h-gain. Reverse annealing (RA) was designed to allow refining a known solution by backward annealing from a classical state representing the solution to a mid-anneal point where a transverse field is present, followed by an ordinary forward anneal, which is hoped to improve on the previous solution. The h-gain (HG) feature stands for time-dependent gain in Hamiltonian linear ($h$) biases and was originally developed to help study freezeout times and phase transitions in spin glasses. Here we apply HG to bias the quantum state in the beginning of the annealing process towards the known solution as in the RA case, but using a different apparatus. We also investigate a hybrid reverse annealing/h-gain schedule, which has a backward phase resembling an RA step and whose forward phase uses the HG idea. To optimize the parameters of the schedules, we employ a Bayesian optimization framework. We test all techniques on a variety of input problems including the weighted Maximum Cut problem and the weighted Maximum Clique problem. Our results show that each technique may dominate the others depending on the input instance, and that the HG technique is a viable alternative to RA for some problems.

quant-ph↗

arXiv2020

Decoy Selection for Protein Structure Prediction Via Extreme Gradient Boosting and Ranking

Identifying one or more biologically-active/native decoys from millions of non-native decoys is one of the major challenges in computational structural biology. The extreme lack of balance in positive and negative samples (native and non-native decoys) in a decoy set makes the problem even more complicated. Consensus methods show varied success in handling the challenge of decoy selection despite some issues associated with clustering large decoy sets and decoy sets that do not show much structural similarity. Recent investigations into energy landscape-based decoy selection approaches show promises. However, lack of generalization over varied test cases remains a bottleneck for these methods. We propose a novel decoy selection method, ML-Select, a machine learning framework that exploits the energy landscape associated with the structure space probed through a template-free decoy generation. The proposed method outperforms both clustering and energy ranking-based methods, all the while consistently offering better performance on varied test-cases. Moreover, ML-Select shows promising results even for the decoy sets consisting of mostly low-quality decoys. ML-Select is a useful method for decoy selection. This work suggests further research in finding more effective ways to adopt machine learning frameworks in achieving robust performance for decoy selection in template-free protein structure prediction.

q-bio.BM↗

arXiv2020

Decomposition algorithms for solving NP-hard problems on a quantum annealer

NP-hard problems such as the maximum clique or minimum vertex cover problems, two of Karp's 21 NP-hard problems, have several applications in computational chemistry, biochemistry and computer network security. Adiabatic quantum annealers can search for the optimum value of such NP-hard optimization problems, given the problem can be embedded on their hardware. However, this is often not possible due to certain limitations of the hardware connectivity structure of the annealer. This paper studies a general framework for a decomposition algorithm for NP-hard graph problems aiming to identify an optimal set of vertices. Our generic algorithm allows us to recursively divide an instance until the generated subproblems can be embedded on the quantum annealer hardware and subsequently solved. The framework is applied to the maximum clique and minimum vertex cover problems, and we propose several pruning and reduction techniques to speed up the recursive decomposition. The performance of both algorithms is assessed in a detailed simulation study.

quant-ph↗

arXiv2019

Peering into the Anneal Process of a Quantum Annealer

Commercial adiabatic quantum annealers have the potential to solve important NP-hard optimization problems efficiently. The newest generation of those machines additionally allows the user to customize the anneal schedule, that is, the schedule with which the anneal fraction is changed from the start to the end of the annealing. In this work we use the aforementioned feature of the D-Wave 2000Q to attempt to monitor how the anneal solution evolves during the anneal process. This process we call slicing: at each time slice during the anneal, we are able to obtain an approximate distribution of anneal solutions. We use our technique to obtain a variety of insights into the D-Wave 2000Q. For example, we observe when individual bits flip during the anneal process and when they stabilize, which allows us to determine the freeze-out point for each qubit individually. We highlight our results using both random QUBO (quadratic unconstrained binary optimization) instances and, for better visualization, instances which we specifically optimize (using our own genetic algorithm) to exhibit a pronounced evolution of its solution during the anneal.

quant-ph↗

arXiv2019

Optimizing the spin reversal transform on the D-Wave 2000Q

Commercial quantum annealers from D-Wave Systems make it possible to obtain approximate solutions of high quality for certain NP-hard problems in nearly constant time. Before solving a problem on D-Wave, several pre-processing methods can be applied, one of them being the so-called spin reversal or gauge transform. The spin reversal transform flips the sign of selected variables and coefficients of the Ising or QUBO (quadratic unconstrained binary optimization) representation of the problem that D-Wave minimizes. The spin reversal transform leaves the ground state of the Ising model invariant, but can average out the biases induced through analog and systematic errors on the device, thus improving the quality of the solution that D-Wave returns. This work investigates the effectiveness of the spin reversal transform for D-Wave 2000Q. We consider two important NP-hard problems, the Maximum Clique and the Minimum Vertex Cover problems, and show on a variety of input problem graphs that using the spin reversal transform can yield substantial improvements in solution quality. In contrast to the native spin reversal built into D-Wave, we consider more general ways to reverse individual spins and we investigate the dependence on the problem type, on the spin reversal probability, and possible advantages of carrying out reversals on the qubit instead of the chain level. Most importantly, for a given individual problem, we use our findings to optimize the spin reversal transform using a genetic optimization algorithm.

quant-ph↗

arXiv2018

Analytical Cost Metrics : Days of Future Past

As we move towards the exascale era, the new architectures must be capable of running the massive computational problems efficiently. Scientists and researchers are continuously investing in tuning the performance of extreme-scale computational problems. These problems arise in almost all areas of computing, ranging from big data analytics, artificial intelligence, search, machine learning, virtual/augmented reality, computer vision, image/signal processing to computational science and bioinformatics. With Moore's law driving the evolution of hardware platforms towards exascale, the dominant performance metric (time efficiency) has now expanded to also incorporate power/energy efficiency. Therefore, the major challenge that we face in computing systems research is: "how to solve massive-scale computational problems in the most time/power/energy efficient manner?" The architectures are constantly evolving making the current performance optimizing strategies less applicable and new strategies to be invented. The solution is for the new architectures, new programming models, and applications to go forward together. Doing this is, however, extremely hard. There are too many design choices in too many dimensions. We propose the following strategy to solve the problem: (i) Models - Develop accurate analytical models (e.g. execution time, energy, silicon area) to predict the cost of executing a given program, and (ii) Complete System Design - Simultaneously optimize all the cost models for the programs (computational problems) to obtain the most time/area/power/energy efficient solution. Such an optimization problem evokes the notion of codesign.

cs.PF↗

arXiv2017

Accelerator Codesign as Non-Linear Optimization

We propose an optimization approach for determining both hardware and software parameters for the efficient implementation of a (family of) applications called dense stencil computations on programmable GPGPUs. We first introduce a simple, analytical model for the silicon area usage of accelerator architectures and a workload characterization of stencil computations. We combine this characterization with a parametric execution time model and formulate a mathematical optimization problem. That problem seeks to maximize a common objective function of 'all the hardware and software parameters'. The solution to this problem, therefore "solves" the codesign problem: simultaneously choosing software-hardware parameters to optimize total performance. We validate this approach by proposing architectural variants of the NVIDIA Maxwell GTX-980 (respectively, Titan X) specifically tuned to a predetermined workload of four common 2D stencils (Heat, Jacobi, Laplacian, and Gradient) and two 3D ones (Heat and Laplacian). Our model predicts that performance would potentially improve by 28% (respectively, 33%) with simple tweaks to the hardware parameters such as adapting coarse and fine-grained parallelism by changing the number of streaming multiprocessors and the number of compute cores each contains. We propose a set of Pareto-optimal design points to exploit the trade-off between performance and silicon area and show that by additionally eliminating GPU caches, we can get a further 2-fold improvement.

cs.AR↗

arXiv2016

Graph-based linear scaling electronic structure theory

We show how graph theory can be combined with quantum theory to calculate the electronic structure of large complex systems. The graph formalism is general and applicable to a broad range of electronic structure methods and materials, including challenging systems such as biomolecules. The methodology combines well-controlled accuracy, low computational cost, and natural low-communication parallelism. This combination addresses substantial shortcomings of linear scaling electronic structure theory, in particular with respect to quantum-based molecular dynamics simulations.

physics.comp-ph↗

arXiv2015

Shortest-Path Queries in Planar Graphs on GPU-Accelerated Architectures

We develop an efficient parallel algorithm for answering shortest-path queries in planar graphs and implement it on a multi-node CPU/GPU clusters. The algorithm uses a divide-and-conquer approach for decomposing the input graph into small and roughly equal subgraphs and constructs a distributed data structure containing shortest distances within each of those subgraphs and between their boundary vertices. For a planar graph with $n$ vertices, that data structure needs $O(n)$ storage per processor and allows queries to be answered in $O(n^{1/4})$ time.

cs.DS↗

arXiv2011

An Approximation Algorithm for Computing Shortest Paths in Weighted 3-d Domains

We present the first polynomial time approximation algorithm for computing shortest paths in weighted three-dimensional domains. Given a polyhedral domain $\D$, consisting of $n$ tetrahedra with positive weights, and a real number $\eps\in(0,1)$, our algorithm constructs paths in $\D$ from a fixed source vertex to all vertices of $\D$, whose costs are at most $1+\eps$ times the costs of (weighted) shortest paths, in $O(\C(\D)\frac{n}{\eps^{2.5}}\log\frac{n}{\eps}\log^3\frac{1}{\eps})$ time, where $\C(\D)$ is a geometric parameter related to the aspect ratios of tetrahedra. The efficiency of the proposed algorithm is based on an in-depth study of the local behavior of geodesic paths and additive Voronoi diagrams in weighted three-dimensional domains, which are of independent interest. The paper extends the results of Aleksandrov, Maheshwari and Sack [JACM 2005] to three dimensions.

cs.CG↗

arXiv2007

A fast multilevel algorithm for graph clustering and community detection

One of the most useful measures of cluster quality is the modularity of a partition, which measures the difference between the number of the edges joining vertices from the same cluster and the expected number of such edges in a random (unstructured) graph. In this paper we show that the problem of finding a partition maximizing the modularity of a given graph G can be reduced to a minimum weighted cut problem on a complete graph with the same vertices as G. We then show that the resulted minimum cut problem can be efficiently solved with existing software for graph partitioning and that our algorithm finds clusterings of a better quality and much faster than the existing clustering algorithms.

physics.data-an↗