Speaker
Description
Although fault-tolerant quantum computation theoretically promises revolutionary benefits for computational science, the hardware development of a scalable fault-tolerant quantum computer is still in its infancy. In this talk, we will demonstrate near-term benefits of quantum or quantum-inspired computation from our two recent works [1-2]. First, using Quantum Alternating Operator Ansatz (QAOA), an algorithm that can be implemented on a near-term quantum device or simulated on a digital computer, we solve a realistic large-scale financial optimization problem experienced by a Thai bank. We show that the presence of the QAOA can improve the expected net profit returned to the bank (in a loan collection problem) by approximately 70%, compared to when the QAOA is absent from the algorithm [1]. Secondly, we discuss the benefits of quantum-inspired approaches in supervised machine learning. In particular, we recast a sentiment analysis problem with a recurrent neural network (RNN), which is difficult to analyze due to intrinsic nonlinearities, as an equivalent problem using a quantum tensor network, which can be efficiently and more easily analyzed on a digital computer. Using an entanglement entropy as a proxy for information propagation, we show that, contrary to a common belief that long-range information propagation is the main source of RNNs’ successes in sentiment analysis, single-layer RNNs harness high expressiveness from the subtle interplay between the information propagation and the word vector embeddings [2]. Our work sheds light on the phenomenology of learning in RNNs, using tools from many-body quantum physics.
References:
[1] https://arxiv.org/abs/2110.15870
[2] https://arxiv.org/abs/2112.08628 (New Journal of Physics, 2022)
