In a recent publication in EPJ Quantum Technology, Le Bin Ho from Tohoku University’s Frontier Institute for Interdisciplinary Sciences has developed a technique called time-dependent stochastic parameter shift in the realm of quantum computing and quantum machine learning. This breakthrough method revolutionizes the estimation of gradients or derivatives of functions, a crucial step in many computational tasks.
Typically, computing derivatives requires dissecting the function and calculating the rate of change over a small interval. But even classical computers cannot keep dividing indefinitely. In contrast, quantum computers can accomplish this task without having to discrete the function. This feature is achievable because quantum computers operate in a realm known as “quantum space,” characterized by periodicity, and no need for endless subdivisions.
One way to illustrate this concept is by comparing the sizes of two elementary schools on a map. To do this, one might print out maps of the schools and then cut them into smaller pieces. After cutting, these pieces can be arranged into a line, with their total length compared (see Figure 1a). However, the pieces may not form a perfect rectangle, leading to inaccuracies. An infinite subdivision would be required to minimize these errors, an impractical solution, even for classical computers.
