Toward a quantum computer that learns from errors

Machine Learning


Dealing with quantum errors

In a concert hall, you can easily hear the sound of an out-of-tune instrument. In the quantum realm, we don’t have that luxury. Measuring qubits collapses the quantum superposition state, as if the very act of listening ruins performance. To preserve quantum information, we instead employ quantum error correction (QEC). It uses redundancy to create “logical qubits” from many physical qubits, and uses specialized parity checks on the physical qubits to digitize analog noise into binary error detection events.

Unfortunately, these bits only indicate that the error occurred somewhere within the limited space-time region of the quantum circuit, not the exact location. It’s like hearing a sour note without knowing exactly which musician played it. To identify potential error locations and calculate the necessary corrections, we utilize QEC decoders such as the neural network decoder AlphaQubit (trained on real data) and the algorithmic decoder Tesseract. If errors are rare enough, these decoders can successfully recover logical quantum information by analyzing the error detection data. However, the decoder leaves an important question unanswered: why these errors occur in the first place.

Some errors arise from the unavoidable interaction of quantum systems with their surrounding environment, leading to decoherence. This ruthless process destroys the macroscopic quantum superposition, effectively turning a quantum computer into a classical computer. This fundamental phenomenon is so pervasive that our familiar classical reality emerges from the underlying quantum laws of nature. While we cannot completely prevent these environmental errors, many other errors are manifestations of inaccurate control calibration or hardware drift, and these deficiencies are within our ability to mitigate.



Source link