Skip to Content

Talk 1: Title: Trainability barriers and opportunities in quantum generative modeling and Talk 2: Title: Reducing molecular electronic Hamiltonian implementation cost for Linear Combination of Unitaries approaches.

Quantum Research Seminars Toronto consist of two 30 min talks about some Quantum Computation topic. Seminars are given by high-level quantum computing researchers with the focus on disseminating their research among other researchers from this field. We encourage to attend researchers regardless of their experience as well as graduate and undergraduate students with particular interest in this field. Basic notions on quantum computing are assumed, but no expertise in any particular subject of this field.

In this 26th series of seminars, the speakers will be Manuel Rudolph of EPFL and Ignacio Loaiza of University of Toronto. Their talks are titled "Trainability barriers and opportunities in quantum generative modeling" and "Reducing molecular electronic Hamiltonian implementation cost for Linear Combination of Unitaries approaches.", respectively.

The event recording, slides and chat history will be published in our Youtube channel and sent to the registered participants.

Looking forward to seeing you all!


Talk 1: Manuel Rudolph

Title: Trainability barriers and opportunities in quantum generative modeling.


In this work, we investigate the barriers to the trainability of quantum generative models posed by barren plateaus and exponential loss concentration. We explore the interplay between explicit and implicit models and losses and show that using implicit generative models (such as quantum circuit-based models) with explicit losses (such as the KL divergence) leads to a new flavour of the barren plateau. In contrast, the Maximum Mean Discrepancy (MMD), which is a popular example of an implicit loss, can be viewed as the expectation value of an observable that is either local and trainable, or global and untrainable depending on the choice of kernel. However, in parallel, we highlight that the local losses required for trainability cannot in general distinguish high-order correlations in the generated data, leading to a fundamental tension between exponential concentration and the emergence of spurious minima. We further propose a new local quantum fidelity-type loss which, by leveraging quantum circuits to estimate the quality of the encoded distribution, is both faithful and enjoys trainability guarantees.

arxiv :

About the Speaker:

Manuel Rudolph is a Ph.D. student at EPFL in Prof. Zoë Holmes' recently formed group. He works on quantum algorithms for near- to mid-term quantum devices, focusing on numerical methods for quantum machine learning approaches such as generative modeling and process learning. Previously, he worked as a Quantum Application Scientist at Zapata Computing, where he developed hybrid algorithms utilizing classical and quantum computing resources. Among other software projects, he also developed the Python visualization library orqviz.

Talk 2: Ignacio Loaiza

Title: Reducing molecular electronic Hamiltonian implementation cost for Linear Combination of Unitaries approaches.


Implementing functions of the electronic structure Hamiltonian H on a quantum computer is a necessary subroutine for many algorithms of interest, such as Quantum Phase Estimation and Ground State Estimation routines. Encoding H as a Linear Combination of Unitaries (LCU) allows for a block-encoding to be implemented on a quantum computer, effectively allowing the efficient implementation of arbitrary polynomials of H in techniques such as qubitization. We will review the LCU framework and how the block-encoding is made, as well as showing some techniques for lowering the cost of implementing functions of H, giving rise to shallower circuits for algorithms of interest.

arxiv :

About the Speaker:

Ignacio Loaiza was born and raised in Mexico City, where he obtained his undergraduate degree in Physics at the Universidad Nacional Autonoma de Mexico (UNAM). He then did his Ph.D. in Theoretical Chemistry at the University of Toronto in Canada, under the supervision of Prof. Artur Izmaylov. During this time, he worked on several topics, including nonadiabatic chemical reactions on metallic surfaces, molecular absorption of solar light, and quantum chemistry methods on quantum computers. He currently holds a Mitacs Elevate postdoctoral fellowship, working jointly with the University of Toronto and Zapata Computing Inc. His research mainly focuses on implementing quantum chemistry methods on quantum computers, particularly on ways to encode the electronic structure Hamiltonian as quantum circuits.

Host: QRST
Event series  CQIQC SeminarsQO/AMO Seminars