Quantum Circuit Training via Generating Functions to Cut Compute Time
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The calculation time for adjusting parameters in quantum machine learning is long due to the complexity of quantum circuits, especially when dealing with deep circuits and a large number of parameters, which can lead to barren plateaus and increased resource consumption.
Innovation Solution
A training device and method that uses a quantum computer to train a machine learning model by optimizing weights of augmentation functions between quantum circuits without updating parameters directly, reducing the need for repeated quantum computations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If quantum machine learning is used to adjust parameters in quantum circuits, then the model can be trained for complex problems, but the calculation time becomes excessively long
Solution Approach 1:
The patent segments the training process into two distinct parts: (1) training the classical neural network parameters using classical computation, and (2) training the quantum circuit parameters using quantum computation. This segmentation allows each part to be optimized independently, reducing the overall calculation time by avoiding quantum computation for all parameters.
Solution Approach 2:
The patent introduces a classical neural network as an intermediary layer between the input data and the quantum circuit. This classical neural network processes and transforms the input data before it enters the quantum circuit, allowing the quantum circuit to focus only on learning the quantum-specific parameters rather than processing raw input data, thereby reducing training time.
2Adaptability or versatility
If the number of parameters in quantum circuits is increased to improve model capacity, then the model can solve more complex problems, but the resource consumption increases
Solution Approach 1:
The patent divides the parameters into classical parameters (processed by classical computers) and quantum parameters (processed by quantum computers). This segmentation allows the system to leverage the strengths of both computational paradigms, reducing the quantum resources needed while maintaining overall model capacity.
Solution Approach 2:
The patent uses a classical neural network to approximate and process parts of the function that would otherwise require quantum computation. This classical copy handles the bulk of the computational workload, allowing the quantum circuit to use fewer parameters and resources while achieving similar or better performance on quantum-specific tasks.
3Manufacturing precision
If deep quantum circuits are used to enhance model performance, then the model can capture complex patterns, but barren plateaus occur making training difficult
Solution Approach 1:
The patent segments the training difficulty by separating classical and quantum parameter training. The classical neural network parameters are trained using standard gradient-based methods that are well-understood and efficient, while the quantum parameters are trained using quantum-specific methods. This segmentation prevents the barren plateau problem from affecting the entire training process.
Solution Approach 2:
The classical neural network acts as an intermediary that preprocessing the input data and providing stabilized gradients to the quantum circuit. This intermediary layer helps mitigate the barren plateau problem by ensuring that the quantum circuit receives well-conditioned inputs and gradients, making training of deep quantum circuits more feasible.
Data Source
AI summary
A training device includes an acquisition unit and a training unit, the acquisition unit acquires a training data set of a machine learning model including a quantum circuit in each of a plurality of layers and a generating function that generates, from an output of a first quantum circuit in a preceding layer in two consecutive layers, an input of a second quantum circuit in a subsequent layer, the training unit determines a value of a parameter included in the generating function by training the machine learning model using a quantum computer that executes calculation of the quantum circuit in each of the plurality of layers and the training data set, and generates the trained machine learning model by setting the value of the parameter in the generating function.