FAQs

TitanQ is a solver platform that can quickly solve very large optimization problems. It is inspired from the mathematics of quantum computing for solving NP-Hard combinatorial optimization problems. For large problems, TitanQ can find optimal and close to optimal solutions very fast by using probabilistic computing. Because its algorithm is probabilistic, it stochastically explores the vast space of probable solutions and is able to return a family of possible solutions to a given problem.

TitanQ can help businesses by offering a pathway to make decisions that are both strategic and data-driven. By pinpointing the most efficient allocation of resources, businesses can significantly cut costs without sacrificing quality or efficiency. Whether it’s enhanced scheduling, route optimization, supply chain management, or something else, there are many pathways where the TitanQ system can improve your business efficiency. This isn't just about minimizing expenses; it's also about maximizing potential revenues. Through techniques like dynamic pricing, companies can adjust their strategies in real-time to better meet market demand, ensuring they're not leaving money on the table. Moreover, TitanQ’s optimization techniques can streamlines operations, reducing bottlenecks and enhancing productivity, which is crucial for staying competitive in today's fast-paced market. Furthermore, as businesses increasingly focus on sustainability, optimization offers a pathway to achieve environmental goals through reduced waste and energy consumption. Ultimately, TitanQ can equip businesses with the tools to make informed, objective decisions, driving improved performance, profitability, and resilience in the face of challenges.

When a problem has more than one solution, it can be difficult to make a decision on which solution is the best. Optimization is a process to find the best solution according to some criteria. For instance, if you are looking for a flight between New York City and Paris, you can find several options. You might make a decision based on price, flight length, number of stops, date, departure time, arrival time, airline reputation, etc. You might consider some of these criteria as constraints, for example dates, or as objectives, for example price. Optimization requires to formulate a mathematical description of the problem to solve, including: · variables: this represents the decisions that one needs to make or the parameters to adjust; · an objective function: this is a mathematical expression that describes the main optimization criteria as a function of the variables; · constraints: constraints describe limits that the solution must not exceed. Constraints can apply to the variables themselves or can be a function of the variables. Constraints are optional in an optimization problem.

While optimization solutions have existed for many years, they are limited by their methods. They tend to struggle to find solutions of problems more than a few hundred variables, and can be limited in their speed for complex use cases. Quantum computers are the next revolution in computing but they are not ready yet with only a few hundreds of qubits available, with limited connectivity and coherence, while also being extremely expensive. Quantum-inspired solutions take inspiration from the exciting methods of quantum computing and develop new methods to solve optimization problems using them. It serves as the next step in optimization and computing and will future-proof your business with the cutting edge of technology.

TitanQ can solve optimization problems that use a mix of binary, integer and continuous variables. It supports problems up to 100,000 binary variables, with linear equality and inequality constraints. It can solve combinatorial problems such as portfolio optimization, vehicle routing, supply chain scheduling, RNA folding and much more, with applications in fields such as logistics, transport, finance, pharmaceutical, energy, and much more.

The TitanQ platform can solve problems at a much larger scale compared to traditional solver platforms (like Gurobi, CPLEX, etc.) and is adding capabilities that are not available in those systems. When we compared 10,000 variable dense problems between the TitanQ platform and Gurobi, we were able to see an >300x speed up in performance with a solution improvement of >1.5% (see our blog, for more details).

We have built an easy to use web API and SDK, described in our documentation here. Send your problem to our system, it will crunch the numbers for you, and you will get an answer back within seconds. Accessing our state of the art system is as simple as a python script, accessible through everything from a raspberry pi, to a high end HPC system, all you need is an internet connection! For customers requiring an on-premise solution, we have solutions available for you too. Please contact us at sales@infinityq.tech for more information.

TitanQ is a cloud-first platform, with the ability to access these revolutionary capabilities through it’s web API and SDK. All you need is an access key to get started. We have special trial programs for startups and small businesses, as well as access for researchers who are interested in testing the platform’s capabilities. Contact sales@infinityq.tech to get more information and access.

The TitanQ system is backed by years of scientific work both within the company and in academia. We build on foundational work in the field of Ising Machines and quantum-inspired methods. Below is a list of relevant papers from InfinityQ employees who have brought their expertise and knowledge to build our revolutionary system: · Ising Model Optimization Problems on a FPGA Accelerated Restricted Boltzmann Machine · Logically Synthesized, Hardware-Accelerated, Restricted Boltzmann Machines for Combinatorial Optimization and Integer Factorization · A Permutational Boltzmann Machine with Parallel Tempering for Solving Combinatorial Optimization Problems Along with this there are many further applications for our models and systems, as outlined in some review papers below: · A Full-Stack View of Probabilistic Computing With p-Bits: Devices, Architectures, and Algorithms · Ising machines as hardware solvers of combinatorial optimization problems

You can get more information on the blog section of our website or by contacting info@infinityq.tech with specific questions.

TitanQ uses quantum-inspired methods, leveraging principles from quantum mechanics to enhance the solving of complex optimization problems. It specifically uses the framework of Ising models to map these problems onto a system of spins with binary states. This approach integrates quantum concepts like superposition—allowing for simultaneous exploration of multiple solutions—and entanglement, which introduces a level of correlation and interaction between variables beyond what classical computing can achieve. Techniques such as simulated annealing, mimicking the cooling process to find minimal energy states, and quantum tunneling, enabling the escape from local minima, are employed to navigate the solution space more effectively, finding optimal solutions that classical methods might miss. This innovative optimization method does not require actual quantum computers but instead runs on classical computers simulating quantum behaviors, making it accessible for current applications. By adopting a hybrid approach that combines the probabilistic nature of quantum mechanics with classical computational power, quantum-inspired optimization offers a robust tool for tackling a wide range of problems—from logistics and finance to machine learning—providing solutions that are often more efficient and practical than those obtained through traditional optimization techniques.

An Ising machine is a specialized computational device designed to solve optimization problems by mapping them onto the Ising model, a mathematical model of ferromagnetism in statistical mechanics. The Ising model represents a system of spins that can be in one of two states, similar to a magnetic dipole pointing up or down. In an Ising machine, these spins are analogous to bits in a computer, but with interactions that mimic the magnetic forces between atoms. The goal is to find the lowest energy configuration of these spins, which corresponds to the optimal solution of the problem that is inputted to the system. Ising machines leverage various physical systems and technologies, including optical, electronic, and quantum systems, to simulate the dynamics of the Ising model and efficiently explore the vast space of possible spin configurations. Ising machines have gained attention for their potential to solve complex optimization problems more efficiently than traditional digital computers, particularly in fields where problems can naturally align with the Ising model structure, such as in logistics, finance, machine learning, and material science. Unlike conventional computers that process information sequentially, Ising machines can explore multiple states simultaneously due to their design, inspired by the parallel processing capabilities seen in natural processes. At InfinityQ, we utilize the parallel nature of the Ising Machine, along with specialized processors, to massively accelerate solutions to many very hard optimization problems. This allows the TitanQ platform to outperform classical computers in many optimization tasks, offering a promising approach to tackling some of the most challenging computational problems faced across various industries.

NP-Hard problems are a category of problems in computer science that are known for their high level of difficulty and exponential scaling with problem size. "NP" stands for "nondeterministic polynomial time," which refers to the class of problems for which a proposed solution can be verified as correct or incorrect in polynomial time, but finding that solution in the first place is not guaranteed to be done within polynomial time. No known algorithm can solve all NP-Hard problems quickly (in polynomial time) for all possible instances of the problem, especially as the size of the problem grows. Ising machines offer a novel approach to tackling NP-Hard problems by leveraging the physics of the Ising model, which involves finding the lowest energy state of a system of interacting spins. The key insight is that many NP-Hard problems can be reformulated as the task of finding the minimum energy configuration of such a system.

Quantum Computing refers to creating a computing using the quantum mechanical principles of superposition, entanglement, and interference. It involves directly linking physical quantum states together, evolving them and then reading out the quantum states at the end of some set of operations. What makes quantum computation so difficult is the preparation of these states, creation of these operators, managing decoherence of the atoms, and controlling sources of noise that could break the quantum system. This usually requires ultra low temperatures, extremely complex electronics, and very precise fabrication technologies. Quantum-Inspired Computing instead does not rely on the an actual set of atoms becoming entangled, but takes inspiration from the exciting physics that quantum computing has shown us. By separating some of the interesting mathematics from the hard to construct, expensive, and difficult to manage system we can get some of the benefits of quantum computing today – running at room temperature, with off the shelf hardware, and without waiting for large quantum computers to be built.

TitanQ runs on high-end graphics processing units (GPU), field programmable gate array (FPGA) and parallel CPU designs. These highly parallel systems are very well suited to the parallel nature of the Ising Model computation, yielding to efficient representations of our systems onto these hardware. By utilizing cloud infrastructure, the TitanQ system also has the ability to scale to many device systems without issue and match the level of workload that our customers have. We plan on scaling the TitanQ platform to allow solutions to problems with millions of variables soon.

While AI systems have become popular in recent years, they struggle very much with structured data and finding the exact solution to problems. If you ask an AI to solve a complex math problem it will most likely get it wrong. On the other hand, systems like the TitanQ platform are purpose built to solve extremely large and complex problems with great precision and certainty. This is how our systems are different from AI. However, stay tuned for how the TitanQ platform can integrate the best of AI with Quantum-Inspired hardware! We hope that the future can marry the ease of use of AI systems, with the amazing computational power of quantum and quantum-inspired technologies.