quantumaudio.schemes

Overview

A Scheme denotes one of the Quantum Audio Representation Methods. This subpackage provides different schemes. The common information for using any scheme is detailed below.

Basic Usage

The core functions of a scheme are encoding and decoding.

  • Encoding: Takes in a digital audio array, performs necessary pre-processing, and prepares a quantum circuit. The quantum circuit can be used to create a state that represents the original digital audio.

  • Decoding: Takes in the quantum circuit measurements, performs necessary post-processing, and reconstructs the original digital audio.

The simplest form of interaction with a Scheme object is to use the encode() and decode() methods, as indicated by the base class quantumaudio.scheme.Scheme.

Detailed Steps

The Encoding and Decoding operations of a scheme can involve several stages that can be manually used for research and debugging purposes. The stages of schemes implemented in the package are listed below.

Encoding

  • Analysis

    • Calculate: Calculates the necessary number of qubits for each quantum register with respect to the data shape, type of scheme, and any user-defined values valid for some schemes. (calculate())

  • Data Pre-Processing

    • Prepare Data: Prepares the data dimension by padding and reshaping. For multi-channel schemes, it also handles the arrangement of samples. (prepare_data())

    • Convert: Converts the data to values suitable for encoding. (convert())

  • Circuit Preparation

    • Initialize Circuit: Initializes the circuit with the calculated number of qubits for each quantum register representing a different aspect of the audio data (i.e. time, value and channel). (initialize_circuit())

    • Value Setting: Encodes or sets the converted values to the circuit. (value_setting())

  • Adding Metadata

    • To keep encode and decode functions independent, key information lost during encoding (e.g., original sample length) is preserved as a Python Dictionary. This can be manually attached to Qiskit circuit’s .metadata attribute or passed separately as argument metadata= in a decode function.

    • The list of metadata keys essential to any specific scheme can be accessed from the scheme’s .keys attribute.

Intermediate

  • Measure

    • Add appropriate classical registers to the encoded circuit for measurement. This can be implemented with a scheme’s measure() method or simply with Qiskit circuit’s measure_all() attribute.

    • By default, encode() method returns measured circuit unless specified measure=False when calling it.

  • Execute

    • The measured and encoded circuit can be executed externally with any provider.

    • By default, the decode() method executes with IBM’s AerSimulator by calling quantumaudio.utils.execute method.

Decoding

  • Decoding Stages

    1. Decode Components: Extracts required components directly from the counts (i.e., a dictionary with the outcome of measurements performed on the quantum circuit). (decode_components())

    2. Undo Conversion: Undoes the data conversion done during encoding. This can be done using the restore() method. (It performs the inverse of convert() method)

    3. Undo Preparation: Undoes the data preparation, such as padding, done during encoding. This can be done manually using numpy slicing and reshape methods. For multi-channel schemes, the arrangement of samples is also restored.

  • Reconstruct Data

    • Takes in a counts dictionary for decoding, combining Decoding Stages 1 and 2. (reconstruct_data())

  • Decode Counts

    • Takes in a counts dictionary for decoding, combining Decoding Stages 1, 2, and 3. It requires metadata to restore the original dimensions of data. (decode_counts())

  • Decode Result

    • Takes in a Qiskit result object for decoding, combining Decoding Stages 1, 2, and 3. It automatically gets additional metadata, such as the original sample length, to undo the padding done at the data preparation stage. (decode_result())

  • Decode

    • Takes in a Qiskit circuit object for decoding, performs measurement (if needed), and default execution, followed by all stages of decoding. (decode())

Contents