/ 最新消息
最新消息

Machine learning, meet quantum computing

2019.05.14|
其他

Today, there is another information processing revolution in its infancy: quantum computing. And that raises an interesting question: is it possible to implement a perceptron on a quantum computer, and if so, how powerful can it be?

Today we get an answer of sorts thanks to the work of Francesco Tacchino and colleagues at the University of Pavia in Italy. These guys have built the world’s first perceptron implemented on a quantum computer and then put it through its paces on some simple image processing tasks.

In its simplest form, a perceptron takes a vector input—a set of numbers—and multiplies it by a weighting vector to produce a single-number output. If this number is above a certain threshold the output is 1, and if it is below the threshold the output is 0.

That has some useful applications. Imagine a pixel array that produces a set of light intensity levels—one for each pixel—when imaging a particular pattern. When this set of numbers is fed into a perceptron, it produces a 1 or 0 output. The goal is to adjust the weighting vector and threshold so that the output is 1 when it sees, say a cat, and 0 in all other cases.

Tacchino and co have repeated Rosenblatt’s early work on a quantum computer. The technology that makes this possible is IBM’s Q-5 “Tenerife” superconducting quantum processor. This is a quantum computer capable of processing five qubits and programmable over the web by anyone who can write a quantum algorithm.

Tacchino and co have created an algorithm that takes a classical vector (like an image) as an input, combines it with a quantum weighting vector, and then produces a 0 or 1 output.

The big advantage of quantum computing is that it allows an exponential increase in the number of dimensions it can process. While a classical perceptron can process an input of   dimensions, a quantum perceptron can process 2dimensions.

Tacchino and co demonstrate this on IBM’s Q-5 processor. Because of the small number of qubits, the processor can handle = 2. This is equivalent to a 2x2 black-and-white image. The researchers then ask: does this image contain horizontal or vertical lines, or a checkerboard pattern?

It turns out that the quantum perceptron can easily classify the patterns in these simple images. “We show that this quantum model of a perceptron can be used as an elementary nonlinear classifier of simple patterns,” say Tacchino and co.

They go on to show how it could be used in more complex patterns, albeit in a way that is limited by the number of qubits the quantum processor can handle.

Ref: arxiv.org/abs/1811.02266 : An Artificial Neuron Implemented on an Actual Quantum Processor