Systems and methods for quantum processing of data using a sparse coded dictionary learned from unlabeled data and supervised learning using encoded labeled data elements
Systems, methods and aspects, and embodiments thereof relate to unsupervised or semi-supervised features learning using a quantum processor. To achieve unsupervised or semi-supervised features learning, the quantum processor is programmed to achieve Hierarchal Deep Learning (referred to as HDL) over one or more data sets. Systems and methods search for, parse, and detect maximally repeating patterns in one or more data sets or across data or data sets. Embodiments and aspects regard using sparse coding to detect maximally repeating patterns in or across data. Examples of sparse coding include L0 and L1 sparse coding. Some implementations may involve appending, incorporating or attaching labels to dictionary elements, or constituent elements of one or more dictionaries. There may be a logical association between label and the element labeled such that the process of unsupervised or semi-supervised feature learning spans both the elements and the incorporated, attached or appended label.
1. A method of automatically labeling data via at least one processor, the at least one processor comprising a quantum processor, the method comprising:
receiving unlabeled data in at least one processor-readable storage medium;
learning, via the quantum processor, a dictionary of dictionary atoms using sparse coding on the received unlabeled data;
receiving labeled data elements in the at least one processor-readable storage medium, each labeled data element which incorporates at least one respective label comprised of at least one respective label element;
preprocessing the labeled data elements to generate preprocessed labeled data elements;
reconstructing, via the at least one processor, the preprocessed labeled data elements using the dictionary to generate encoded labeled data elements;
executing, via the at least one processor, a supervised learning process using the encoded labeled data elements to produce at least one of a classifier or a label assigner; and
storing the produced at least one classifier or label assigner in the at least one processor-readable storage medium.
2. The method of claim 1 wherein the labeled data elements comprise labeled image data elements, and preprocessing the labeled data elements to generate preprocessed labeled data elements comprises at least one of:
normalizing at least one of a contrast or a brightness of the image data elements;
whitening the image data elements;
applying zero phase component analysis (ZCA) whitening to the image data elements; or
reducing a dimensionality of the image data elements.
3. The method of claim 1 wherein executing, via the at least one processor, a supervised learning process comprises performing at least one of a perceptron algorithm, a k nearest neighbors (kNN) algorithm, or a linear support vector machine (SVM) with L1 and L2 loss algorithm.
4. The method of claim 1 wherein receiving labeled data elements in the at least one processor-readable storage medium comprises receiving labeled image data elements, each labeled image data element incorporates at least one respective label comprised of at least one respective image label element.
5. The method of claim 1 wherein receiving labeled data elements in the at least one processor-readable storage medium comprises receiving labeled data elements each of a specific type or format of data, and each labeled data element is of the same specific type or format of data as the received respective label element.
6. The method of claim 1 wherein learning, via the quantum processor, a dictionary of dictionary atoms using sparse coding on the received unlabeled data includes:
formulating an objective function based on the received unlabeled data via the at least one processor; and
attempting to minimize the objective function via the quantum processor.
7. The method of claim 6 wherein formulating an objective function based on the received unlabeled data via at least one processor includes formulating the objective function based on the received unlabeled data via the at least one processor, the objective function comprising a regularization term governed by an L0-norm form.
8. The method of claim 7 , further comprising:
casting a set of weights in the objective function as Boolean variables via at least one processor to generate Boolean weights;
setting a set of values for the dictionary via the at least one processor; and
attempting to optimize the objective function for a set of values for the Boolean weights based on the set of values for the dictionary via the quantum processor.
9. The method of claim 8 wherein attempting to optimize the objective function for a set of values for the Boolean weights includes mapping the objective function to a quadratic unconstrained binary optimization (“QUBO”) problem and using the quantum processor to attempt to at least approximately minimize the QUBO problem.
10. The method of claim 9 wherein using the quantum processor to attempt to at least approximately minimize the QUBO problem includes using the quantum processor to perform at least one of adiabatic quantum computation or quantum annealing.
11. The method of claim 1 wherein the labeled data elements comprise labeled image data elements, and preprocessing the labeled data elements to generate preprocessed labeled data elements comprises normalizing at least one of a contrast or a brightness of the image data elements.
12. The method of claim 1 wherein the labeled data elements comprise labeled image data elements, and preprocessing the labeled data elements to generate preprocessed labeled data elements comprises whitening the image data elements.
13. The method of claim 1 wherein the labeled data elements comprise labeled image data elements, and preprocessing the labeled data elements to generate preprocessed labeled data elements comprises applying zero phase component analysis (ZCA).
14. The method of claim 1 wherein the labeled data elements comprise labeled image data elements, and preprocessing the labeled data elements comprises reducing a dimensionality of the image data elements.
15. The method of claim 1 wherein executing, via at least one processor, a supervised learning process comprises performing a perceptron algorithm.
16. The method of claim 1 wherein executing, via at least one processor, a supervised learning process comprises performing a k nearest neighbors (kNN) algorithm.
17. The method of claim 1 wherein executing, via at least one processor, a supervised learning process comprises performing a linear support vector machine (SVM) with L1 and L2 loss algorithm.