A central question for active learning (AL) is: what is the optimal selection? Defining optimality by classifier loss produces a new characterisation of optimal AL behaviour, by treating expected loss reduction as a statistical target for estimation.
This target forms the basis of model retraining improvement (MRI), a novel approach providing a statistical estimation framework for AL. This framework is constructed to address the central question of AL optimality, and to motivate the design of estimation algorithms. MRI allows the exploration of optimal AL behaviour, and the examination of AL heuristics, showing precisely how they make sub-optimal selections. The abstract formulation of MRI is used to provide a new guarantee for AL, that an unbiased MRI estimator should outperform random selection. This MRI framework reveals intricate estimation issues that in turn motivate the construction of new statistical AL algorithms. One new algorithm in particular performs strongly in a large-scale experimental study, compared to standard AL methods. This competitive performance suggests that practical efforts to minimise estimation bias may be important for AL applications.
In many classification problems unlabelled data is abundant and a subset can be chosen for labelling. This defines the context of active learning (AL), where methods systematically select that subset, to improve a classifier by retraining. Given a cl
assification problem, and a classifier trained on a small number of labelled examples, consider the selection of a single further example. This example will be labelled by the oracle and then used to retrain the classifier. This example selection raises a central question: given a fully specified stochastic description of the classification problem, which example is the optimal selection? If optimality is defined in terms of loss, this definition directly produces expected loss reduction (ELR), a central quantity whose maximum yields the optimal example selection. This work presents a new theoretical approach to AL, example quality, which defines optimal AL behaviour in terms of ELR. Once optimal AL behaviour is defined mathematically, reasoning about this abstraction provides insights into AL. In a theoretical context the optimal selection is compared to existing AL methods, showing that heuristics can make sub-optimal selections. Algorithms are constructed to estimate example quality directly. A large-scale experimental study shows these algorithms to be competitive with standard AL methods.
We demonstrate the coherent transduction of quantum noise reduction, or squeezed light, by Ag localized surface plasmons (LSPs). Squeezed light, generated through four-wave-mixing in Rb vapor, is coupled to a Ag nanohole array designed to exhibit LSP
-mediated extraordinary-optical transmission (EOT) spectrally coincident with the squeezed light source at 795 nm. We demonstrate that quantum noise reduction as a function of transmission is found to match closely with linear attenuation models, thus demonstrating that the photon-LSP-photon transduction process is coherent near the LSP resonance.
We present results of a bright polarization-entangled photon source operating at 1552 nm via type-II collinear degenerate spontaneous parametric down-conversion in a periodically poled potassium titanyl phosphate crystal. We report a conservative inf
erred pair generation rate of 123,000 pairs/s/mW into collection modes. Minimization of spectral and spatial entanglement was achieved by group velocity matching the pump, signal and idler modes and through properly focusing the pump beam. By utilizing a pair of calcite beam displacers, we are able to overlap photons from adjacent down-conversion processes to obtain polarization-entanglement visibility of 94.7 +/- 1.1% with accidentals subtracted.
