An initialization strategy for addressing barren plateaus in parametrized quantum circuits

  title={An initialization strategy for addressing barren plateaus in parametrized quantum circuits},
  author={Edward Grant and Leonard Wossnig and Mateusz Ostaszewski and Marcello Benedetti},
  journal={arXiv: Quantum Physics},
Parametrized quantum circuits initialized with random initial parameter values are characterized by barren plateaus where the gradient becomes exponentially small in the number of qubits. In this technical note we theoretically motivate and empirically validate an initialization strategy which can resolve the barren plateau problem for practical applications. The technique involves randomly selecting some of the initial parameter values, then choosing the remaining values so that the circuit is… Expand
Absence of Barren Plateaus in Quantum Convolutional Neural Networks
This work rigorously analyze the gradient scaling for the parameters in the QCNN architecture and finds that the variance of the gradient vanishes no faster than polynomially, implying that QCNNs do not exhibit barren plateaus. Expand
Quantum compilation and circuit optimisation via energy dissipation
We describe a method for automatically recompiling a quantum circuit A into a target circuit B, with the goal that both circuits have the same action on a specific input i.e. A|in> = B|in>. This isExpand
Effect of barren plateaus on gradient-free optimization
It is shown that gradient-free optimizers do not solve the barren plateau problem, and the main result proves that cost function differences, which are the basis for making decisions in a gradient- free optimization, are exponentially suppressed in a barren plateau. Expand
Entanglement devised barren plateau mitigation
Hybrid quantum-classical variational algorithms are one of the most propitious implementations of quantum computing on near-term devices, offering classical machine learning support to quantum scaleExpand
Quantum Generative Training Using R\'enyi Divergences
Quantum neural networks (QNNs) are a framework for creating quantum algorithms that promises to combine the speedups of quantum computation with the widespread successes of machine learning. A majorExpand
Capacity and quantum geometry of parametrized quantum circuits
The capacity and trainability of parametrized quantum circuits are evaluated using the geometric structure of the parameter space via the effective quantum dimension, which reveals the expressive power of circuits in general as well as of particular initialization strategies. Expand
Layerwise learning for quantum neural networks
This work investigates a layerwise learning strategy for parametrized quantum circuits and shows that when considering sampling noise, this strategy can help avoid the problem of barren plateaus of the error surface due to the low depth of circuits, low number of parameters trained in one step, and larger magnitude of gradients compared to training the full circuit. Expand
Barren Plateaus Preclude Learning Scramblers.
A no-go theorem for learning an unknown scrambling process with QML is proved, showing that it is highly probable for any variational Ansatz to have a barren plateau landscape, i.e., cost gradients that vanish exponentially in the system size. Expand
On barren plateaus and cost function locality in variational quantum algorithms
A lower bound on the variance of the gradient is derived, which depends mainly on the width of the circuit causal cone of each term in the Pauli decomposition of the cost function. Expand
The Born supremacy: quantum advantage and training of an Ising Born machine
This work defines a subset of a class of quantum circuits known as Born machines based on Ising Hamiltonians and shows that the circuits encountered during gradient-based training cannot be efficiently sampled from classically up to multiplicative error in the worst case. Expand


Barren plateaus in quantum neural network training landscapes
It is shown that for a wide class of reasonable parameterized quantum circuits, the probability that the gradient along any reasonable direction is non-zero to some fixed precision is exponentially small as a function of the number of qubits. Expand
Universal discriminative quantum neural networks
This work trains near-term quantum circuits to classify data represented by non-orthogonal quantum probability distributions using the Adam stochastic optimization algorithm, and provides an example of quantum machine learning for a task that has inherently no classical analogue. Expand
Hardware-efficient variational quantum eigensolver for small molecules and quantum magnets
The experimental optimization of Hamiltonian problems with up to six qubits and more than one hundred Pauli terms is demonstrated, determining the ground-state energy for molecules of increasing size, up to BeH2. Expand
Hierarchical quantum classifiers
It is shown how quantum algorithms based on two tensor network structures can be used to classify both classical and quantum data and may enable classification of two-dimensional images and entangled quantum data more efficiently than is possible with classical methods. Expand
A variational eigenvalue solver on a photonic quantum processor
The proposed approach drastically reduces the coherence time requirements and combines this method with a new approach to state preparation based on ansätze and classical optimization, enhancing the potential of quantum resources available today and in the near future. Expand
Quantum t-designs: t-wise Independence in the Quantum World
  • A. Ambainis, J. Emerson
  • Computer Science, Physics
  • Twenty-Second Annual IEEE Conference on Computational Complexity (CCC'07)
  • 2007
It is shown that an approximate 4-design provides a derandomization of the statedistinction problem considered by Sen (quant-ph/0512085), which is relevant to solving certain instances of the hidden subgroup problem. Expand
Adam: A Method for Stochastic Optimization
This work introduces Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments, and provides a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Expand
Symbolic integration with respect to the Haar measure on the unitary group in Mathematica
A number of special cases which can be used to optimize the calcu-lation speed for some classes of integrals of the IntU package for Mathematica computer algebra system are described. Expand
Circuit-centric quantum classifiers
A machine learning design is developed to train a quantum circuit specialized in solving a classification problem. In addition to discussing the training method and effect of noise, it is shown thatExpand
Unitary 2-designs, variational quantum eigensolvers, and barren plateaus
  • https://qitheory.,
  • 2019