It took me a surprisingly long time to gather the necessary ingredients for the edge quantum computing recipe. I had to visit multiple libraries and talk to much more experienced chefs. From what I understand, the scarcity of resources is due to the complexity of the field. Edge quantum computing is still emerging, and many underlying tools have yet to be discovered. Moreover, its final deployment will require an interdisciplinary effort and knowledge, rarely found in a single place. Nevertheless, today we make a plan and list the ingredients. Tomorrow, we start cooking.
Ingredients
Below, I list the main ingredients that will be reflected in upcoming posts. Once a post is ready, I’ll link it here.
Physics:
Probing quantum systems at nanosecond timescales gives us the opportunity to study quantum phenomena in real time. Ultimately, we aim to shed new light on certain aspects of the quantum world and guide toward answers to fundamental questions that have puzzled humanity for the past century. A deeper understanding of these basics will also help in designing tools that better exploit the quantum nature of systems.
– Quantum Superposition: This is needed to achieve apparent “parallel” computation and, more generally, to benefit from using quantum systems. We will discuss the meaning of superposition and its difference from classical probability distributions.
– Measurement Problem: Bravely, we will introduce arguably the most controversial topic in quantum mechanics—the act of turning probabilities into one observed outcome. In doing so, we’ll explore different types of measurements and their implications on quantum systems.
– Decoherence: “If you know your enemy and know yourself, you need not fear the result of a hundred battles” – Sun Tzu. Decoherence is our enemy. We will explore its relation to the measurement problem and distinguish the different ways quantum correlations can be lost.
Engineering:
Edge quantum computing is not only about fundamental physics but also about cutting-edge engineering tools, which are still in the development phase. Understanding the current state of the art and future directions will be essential.
– Control of Quantum Systems: Depending on the hardware and qubit encoding, the control of quantum systems can occur on different timescales. Since control fields are often electromagnetic, faster operations typically increase decoherence due to environmental noise. We will discuss state-of-the-art methods for realizing and optimizing qubit control.
– FPGA: Field Programmable Gate Arrays (FPGAs) are a natural element of the quantum edge computing pipeline. They can be leveraged to significantly lower readout, processing, and control times. We’ll examine the current state of the art and future directions for FPGAs in quantum computing.
Machine Learning:
The vision of edge computing includes an intelligent agent that can make informed decisions based on data from the quantum computer. We will track the state-of-the-art developments in machine learning algorithms that can first be tested offline and later deployed in edge quantum computing.
– Reinforcement Learning: The communication between the quantum computer and the controller happens in well-defined steps—sending a pulse (action) and receiving a bit (state/belief)—which is the natural language of RL. In principle, the agent can learn optimal control strategies based on feedback from the quantum computer.
– Recurrent Neural Networks (RNNs) and Bayesian methods: Quantum computers are dynamical systems that can be described by time series of control pulses and qubit states. RNNs are well-suited to model such systems and predict future states based on past observations. Additionally when model is available, Bayesian methods can be used to infer the underlying physics and use adaptive strategy.
– Physics-Informed Machine Learning: Since quantum computers are physical systems governed by the laws of physics, machine learning algorithms can be used to infer underlying physics and optimize control strategies based on physical models.