Quantum Neural Networks / Parameterized Quantum Models Model - f Takes in data - x Depends on Parameters - theta f(x, theta) 2 major design choices 1. Model Family 2. Parameterize the model - Find opimal parameters for robust predictions How do we measure the power of a model? 1. Count the parameters 2. Vapnik-Chernovenskis (VC) dimension 1. Measure model capacity, expressibility and complexity 2. Can be hard to compute, generalisation bounds are lose (undesirable) 3. Effective Dimension 1. Aims to capture relevance and generalizes integer dimensions - what is relevant and being used in the model 2. scale dependent measure - ML Alogrithms on x Axis - Quantum - Classical - Type of Data - Quantum - Classical Analyse your quantum data - Quantum Circuit - Apply gates - Get quantum states Get classical data from your device then do ML on that data to understand quantum Gate form is when it is in a state vector Quantum Data is in Superposed State - Manipulation being within quantum mechanics to be able to go back into the quantum domain from the classical domain - How to operate with quantum state vectors - - xx Qiml vs classical - ml as a block box. - feed data - get output box is diff q(input, parameters) = output) f(input, parameters) = output) nuerons = numbers of connections optical encoding for 1 neuron to 1 qubit quantum variational model - gates model weights that are co-related we construct the quantun not independent - input is a 3 qubit system - 3 dimensions qml vs qml