We prove that using global observables to train the matrix product state ansatz results in the vanishing of all partial derivatives, also known as barren plateaus, while using local observables avoids this. This ansatz is widely used in quantum machine learning for learning weakly entangled state approximations. Additionally, we empirically demonstrate that in many cases, the objective function is an inner product of almost sparse operators, highlighting the potential for classically simulating such a learning problem with few quantum resources. All our results are experimentally validated across various scenarios.

Algorithms for processing data in short-time batches are critical for both online and offline processing of streamed and large data respectively due to the quadratic relation between signal length and computational cost of convolution-based processing schemes. Whilst quantum analogs to some digital signal processing algorithms have been discovered, including the quantum Fourier transform (QFT), there has been no development of short-time processing techniques in the quantum domain. In this manuscript, we introduce the short-time QFT (STQFT) processing technique to bridge this gap in research. We develop a novel overlap-add reconstruction technique in the quantum domain using a permutation gate to combine subsequent windows. With this in mind, we discuss convolution under our novel STQFT processing scheme. We demonstrate filtering in the quantum Fourier domain with a filter stored in a quantum register as well as in a block encoded unitary gate. Throughout the paper, we elaborate upon implementation details such as applying DC offsets to input signals, skipping input data frames whenever necessary, the use of overlap-save as a reconstruction technique and mitigating time-varying scaling due to normalization of the windowed input data and filters.

Variational quantum algorithms (VQAs) are the quantum analog of classical neural networks (NNs). A VQA consists of a parameterized quantum circuit (PQC) which is composed of multiple layers of ansatzes (simpler PQCs, which are an analogy of NN layers) that differ only in selections of parameters. Previous work has identified the alternating layered ansatz as potentially a new standard ansatz in near-term quantum computing. Indeed, shallow alternating layered VQAs are easy to implement and have been shown to be both trainable and expressive. In this work, we introduce a training algorithm with an exponential reduction in training cost of such VQAs. Moreover, our algorithm uses classical shadows of quantum input data, and can hence be run on a classical computer with rigorous performance guarantees. We demonstrate 2-3 orders of magnitude improvement in the training cost using our algorithm for the example problems of finding state preparation circuits and the quantum autoencoder.