Mimicking human brains in the quantum world

Quantum computers that learn like humans have the potential to solve challenging problems ranging from drug design to quantum encryption. Image credit: Pixabay. 

Once a scientific curiosity, quantum computers have gained attention in recent years for their phenomenal computational abilities that may be used to design more effective drugs, securely transmit information, or predict the weather.

Harnessing the power of quantum computing and artificial intelligence, NTU scientists have designed a prototype quantum computer that learns like a human, paving the way for even more powerful quantum computers with unprecedented problem-solving capabilities.

Equipped with an artificial brain, the new quantum computer learned quickly to perform complex computations on its own.   

Enabling a quantum computer to learn from experience

Unlike conventional computer chips that store information in states of 1’s and 0’s, information in the quantum computing world is stored as states in quantum particles known as qubits. Using computer simulations, the scientists at NTU’s School of Physical and Mathematical Sciences developed and trained an artificial neural network, which is a processor containing a network of computational units that mimics how cells are connected in the brain, to analyse data stored in quantum particles. 

Like a brain, the artificial neural network in the quantum computer can learn to recognise patterns in quantum data. Image credit: Sanjib Ghosh. 

Quantum computers process information at ultrafast speeds, thanks to entangled qubits, in which changing the state of one qubit instantaneously changes the state of another qubit in a predictable way. As quantum entanglement is an essential feature of quantum computing, the researchers trained their quantum computer to recognise entangled states so that it could identify quantum entanglement.

The scientists input examples of quantum data and their expected entangled states into the computer, and trained it to identify quantum entanglement, much like teaching a child how to recognise objects using flashcards.

After only 200 examples, fewer than the thousands of examples typically required to train most artificially intelligent computers, the prototype quantum computer was able to identify quantum entanglement when presented with a quantum entangled state that it had not encountered before.

“For quantum computers to function optimally, it is crucial that they generate the correct quantum entangled states. Thus, our quantum computer can be used as a tool to develop other quantum computers,” said Asst Prof Timothy Liew who led the research.

“Besides identifying quantum entanglement, our quantum computer can also potentially learn to perform notoriously difficult mathematical calculations used in applications such as quantum cryptography in which information is encrypted in quantum particles and securely exchanged,” added Asst Prof Liew. 

The researchers envision that their quantum computer will open new possibilities for quantum computers to solve complex problems that may revolutionise everyday life.

 

More details of this research can be found in “Reconstructing quantum states with quantum reservoir networks”, published in IEEE Transactions on Neural Networks and Learning Systems (2020). DOI: 10.1109/TNNLS.2020.3009716; “Quantum reservoir processing”, published in npj Quantum Information (2019). DOI: 10.1038/s41534-019-0149-8; and “Quantum neuromorphic platform for quantum state preparation”, published in Physical Review Letters (2019). DOI: 10.1103/PhysRevLett.123.260404.