Thursday, August 31

Quantum Computing 10 Times More Efficient


In the world of quantum error correction, an underdog is coming for the king.

Last week, new simulations from two groups reported that a rising class of quantum error-correcting codes is more efficient by an order of magnitude than the current gold standard, known as the surface code. 

The codes all work by transforming a horde of error-prone qubits into a much smaller band of “protected” qubits that rarely make mistakes. But in the two simulations, low-density parity check — or LDPC — codes could make protected qubits out of 10 to 15 times fewer raw qubits than the surface code. 

Neither group has implemented these simulated leaps in actual hardware, but the experimental blueprints suggest that these codes, or codes like them, could hasten the arrival of more capable quantum devices.

“It really looks like it’s coming to fruition,” said Daniel Gottesman of the University of Maryland, who studies LDPC codes but was not involved in the recent studies. “These [codes] could be practical things that can greatly improve our ability to make quantum computers.”

Classical computers run on bits that rarely misfire. But the particle-like objects — qubits — that power quantum computers lose their quantum mojo when just about anything jostles them out of their delicate state. 

To coax future qubits into usefulness, researchers plan to use quantum error correction, the practice of using extra qubits to redundantly encode information. It’s similar in spirit to protecting a message from static by speaking each word twice, spreading out the information among more characters.

The Canonical King
In 1998, Alexei Kitaev of the California Institute of Technology and Sergey Bravyi, then of the Landau Institute for Theoretical Physics in Russia, introduced the quantum error-correcting surface code. It organizes qubits into a square grid and executes something like a game of Minesweeper: Each qubit connects to four neighbors, so checking designated helper qubits allows you to discreetly snoop on four data-carrying qubits. 

Depending on whether the check returns a 0 or a 1, you can infer whether some of the neighbors have erred. By checking around the board, you can deduce where the errors are and fix them.  READ MORE...

No comments:

Post a Comment