Quantum Techniques in Machine Learning (QTML) is an annual international conference focusing on the interdisciplinary field of quantum technology and machine learning. The goal of the conference is to gather leading academic researchers and industry players to interact through a series of scientific talks focussed on the interplay between machine learning and quantum physics.
In this tutorial you will learn about mathematical formulations of computational learning theory and its generalization to problems in quantum information and computing. We will discuss basic definitions and problems investigated in this area. This will help you to start your journey in this exciting and high-profile field.
Shallow quantum circuits lie at the forefront of modern experimental capabilities and theoretical understanding. In this work, we present the first computationally-efficient algorithm for average-case learning of a class of shallow circuits with many-qubit gates. Namely, we provide a quasipolynomial sample and time complexity algorithm for learning 1/poly(n)-approximate unitaries of QAC^0 circuits, i.e., constant-depth circuits with arbitrary single-qubit gates and polynomially many CZ gates of un-bounded width. Note that QAC^0 implements circuits requiring linear-depth in the 1D geometry and
logarithmic-depth in the all-to-all-connected geometry of constant-width gates. Furthermore, leveraging our learned QAC0 unitaries, we provide an efficient algorithm for circuit synthesis of poly-logarithmic depth QAC circuits, making progress towards a quasipolynomial proper-learning algorithm for QAC0.
Presentations from Industry
This presentation will give a brief overview of CSIRO (Australia's National Science Agency) and our work in quantum technologies.
A fundamental problem in quantum many-body physics is that of finding ground states of local Hamiltonians. A number of recent works gave provably efficient machine learning (ML) algorithms for learning ground states. Specifically, [Huang et al. Science 2022], introduced an approach for learning properties of the ground state of an n-qubit gapped local Hamiltonian H from only poly(n) data points sampled from Hamiltonians in the same phase of matter. This was subsequently improved by [Lewis et al. Nature Communications 2024], to log(n) samples when the geometry of the n-qubit system is known. In this work, we introduce two approaches that achieve a constant sample complexity, independent of system size n, for learning ground state properties. Our first algorithm consists of a simple modification of the ML model used by Lewis et al. and applies to a property of interest known beforehand. Our second algorithm, which applies even if a description of the property is not known, is a deep neural network model. While empirical results showing the performance of neural networks have been demonstrated, to our knowledge, this is the first rigorous sample complexity bound on a neural network model for predicting ground state properties. We also perform numerical experiments that confirm the improved scaling of our approach compared to earlier results.
We present an entropy based method for adaptive magnetometry, focusing on myopic entropy optimisation of NV Ramsey measurements.
We design a quantum method for classical information compression that exploits the hidden subgroup quantum algorithm. We consider sequence data in a database with a priori unknown symmetries of the hidden subgroup type. We prove that data with a given group structure can be compressed with the same query complexity as the hidden subgroup problem, which is exponentially faster than the best-known classical algorithms. We moreover design a quantum algorithm that variationally finds the group structure and uses it to compress the data. There is an encoder and a decoder, along the paradigm of quantum autoencoders. After the training, the encoder outputs a compressed data string and a description of the hidden subgroup symmetry, from which the input data can be recovered by the decoder. In illustrative examples, our algorithm outperforms the classical autoencoder on the mean squared value of test data. This classical-quantum separation in information compression capability has thermodynamical significance: the free energy assigned by a quantum agent to a system can be much higher than that of a classical agent. Taken together, our results show that a possible application of quantum computers is to efficiently compress certain types of data that cannot be efficiently compressed by current methods using classical computers.