-
Metrics and methods for a systematic comparison of fairness-aware machine learning algorithms
Authors:
Gareth P. Jones,
James M. Hickey,
Pietro G. Di Stefano,
Charanpal Dhanjal,
Laura C. Stoddart,
Vlasios Vasileiou
Abstract:
Understanding and removing bias from the decisions made by machine learning models is essential to avoid discrimination against unprivileged groups. Despite recent progress in algorithmic fairness, there is still no clear answer as to which bias-mitigation approaches are most effective. Evaluation strategies are typically use-case specific, rely on data with unclear bias, and employ a fixed policy…
▽ More
Understanding and removing bias from the decisions made by machine learning models is essential to avoid discrimination against unprivileged groups. Despite recent progress in algorithmic fairness, there is still no clear answer as to which bias-mitigation approaches are most effective. Evaluation strategies are typically use-case specific, rely on data with unclear bias, and employ a fixed policy to convert model outputs to decision outcomes. To address these problems, we performed a systematic comparison of a number of popular fairness algorithms applicable to supervised classification. Our study is the most comprehensive of its kind. It utilizes three real and four synthetic datasets, and two different ways of converting model outputs to decisions. It considers fairness, predictive-performance, calibration quality, and speed of 28 different modelling pipelines, corresponding to both fairness-unaware and fairness-aware algorithms. We found that fairness-unaware algorithms typically fail to produce adequately fair models and that the simplest algorithms are not necessarily the fairest ones. We also found that fairness-aware algorithms can induce fairness without material drops in predictive power. Finally, we found that dataset idiosyncracies (e.g., degree of intrinsic unfairness, nature of correlations) do affect the performance of fairness-aware approaches. Our results allow the practitioner to narrow down the approach(es) they would like to adopt without having to know in advance their fairness requirements.
△ Less
Submitted 8 October, 2020;
originally announced October 2020.
-
Fairness by Explicability and Adversarial SHAP Learning
Authors:
James M. Hickey,
Pietro G. Di Stefano,
Vlasios Vasileiou
Abstract:
The ability to understand and trust the fairness of model predictions, particularly when considering the outcomes of unprivileged groups, is critical to the deployment and adoption of machine learning systems. SHAP values provide a unified framework for interpreting model predictions and feature attribution but do not address the problem of fairness directly. In this work, we propose a new definit…
▽ More
The ability to understand and trust the fairness of model predictions, particularly when considering the outcomes of unprivileged groups, is critical to the deployment and adoption of machine learning systems. SHAP values provide a unified framework for interpreting model predictions and feature attribution but do not address the problem of fairness directly. In this work, we propose a new definition of fairness that emphasises the role of an external auditor and model explicability. To satisfy this definition, we develop a framework for mitigating model bias using regularizations constructed from the SHAP values of an adversarial surrogate model. We focus on the binary classification task with a single unprivileged group and link our fairness explicability constraints to classical statistical fairness metrics. We demonstrate our approaches using gradient and adaptive boosting on: a synthetic dataset, the UCI Adult (Census) dataset and a real-world credit scoring dataset. The models produced were fairer and performant.
△ Less
Submitted 26 June, 2020; v1 submitted 11 March, 2020;
originally announced March 2020.
-
Counterfactual fairness: removing direct effects through regularization
Authors:
Pietro G. Di Stefano,
James M. Hickey,
Vlasios Vasileiou
Abstract:
Building machine learning models that are fair with respect to an unprivileged group is a topical problem. Modern fairness-aware algorithms often ignore causal effects and enforce fairness through modifications applicable to only a subset of machine learning models. In this work, we propose a new definition of fairness that incorporates causality through the Controlled Direct Effect (CDE). We deve…
▽ More
Building machine learning models that are fair with respect to an unprivileged group is a topical problem. Modern fairness-aware algorithms often ignore causal effects and enforce fairness through modifications applicable to only a subset of machine learning models. In this work, we propose a new definition of fairness that incorporates causality through the Controlled Direct Effect (CDE). We develop regularizations to tackle classical fairness measures and present a causal regularization that satisfies our new fairness definition by removing the impact of unprivileged group variables on the model outcomes as measured by the CDE. These regularizations are applicable to any model trained using by iteratively minimizing a loss through differentiation. We demonstrate our approaches using both gradient boosting and logistic regression on: a synthetic dataset, the UCI Adult (Census) Dataset, and a real-world credit-risk dataset. Our results were found to mitigate unfairness from the predictions with small reductions in model performance.
△ Less
Submitted 26 February, 2020; v1 submitted 25 February, 2020;
originally announced February 2020.
-
Ultrastrong coupling probed by Coherent Population Transfer
Authors:
G. Falci,
A. Ridolfo,
P. G. Di Stefano,
E. Paladino
Abstract:
Light-matter interaction, and the understanding of the fundamental physics behind, is the scenario of emerging quantum technologies. Solid state devices allow the exploration of new regimes where ultrastrong coupling (USC) strengths are comparable to subsystem energies, and new exotic phenomena like quantum phase transitions and ground-state entanglement occur. While experiments so far provided on…
▽ More
Light-matter interaction, and the understanding of the fundamental physics behind, is the scenario of emerging quantum technologies. Solid state devices allow the exploration of new regimes where ultrastrong coupling (USC) strengths are comparable to subsystem energies, and new exotic phenomena like quantum phase transitions and ground-state entanglement occur. While experiments so far provided only spectroscopic evidence of USC, we propose a new dynamical protocol for detecting virtual photon pairs in the dressed eigenstates. This is the fingerprint of the violated conservation of the number of excitations, which heralds the symmetry broken by USC. We show that in flux-based superconducting architectures this photon production channel can be coherenly amplified by Stimulated Raman Adiabatic Passage (STIRAP). This provides a unique tool for an unambiguous dynamical detection of USC in present day hardware. Implementing this protocol would provide a benchmark for control of the dynamics of USC architectures, in view of applications to quantum information and microwave quantum photonics.
△ Less
Submitted 27 March, 2019; v1 submitted 2 August, 2017;
originally announced August 2017.
-
Advances in quantum control of three-level superconducting circuit architectures
Authors:
G. Falci,
P. G. Di Stefano,
A. Ridolfo,
A. D'Arrigo,
G. S. Paraoanu,
E. Paladino
Abstract:
Advanced control in Lambda ($Λ$) scheme of a solid state architecture of artificial atoms and quantized modes would allow the translation to the solid-state realm of a whole class of phenomena from quantum optics, thus exploiting new physics emerging in larger integrated quantum networks and for stronger couplings. However control solid-state devices has constraints coming from selection rules, du…
▽ More
Advanced control in Lambda ($Λ$) scheme of a solid state architecture of artificial atoms and quantized modes would allow the translation to the solid-state realm of a whole class of phenomena from quantum optics, thus exploiting new physics emerging in larger integrated quantum networks and for stronger couplings. However control solid-state devices has constraints coming from selection rules, due to symmetries which on the other hand yield protection from decoherence, and from design issues, for instance that coupling to microwave cavities is not directly switchable. We present two new schemes for the $Λ$-STIRAP control problem with the constraint of one or two classical driving fields being always-on. We show how these protocols are converted to apply to circuit-QED architectures. We finally illustrate an application to coherent spectroscopy of the so called ultrastrong atom-cavity coupling regime.
△ Less
Submitted 3 April, 2017;
originally announced April 2017.
-
Non-equilibrium thermodynamics of continuously measured quantum systems: a circuit-QED implementation
Authors:
P. G. Di Stefano,
J. J. Alonso,
E. Lutz,
G. Falci,
M. Paternostro
Abstract:
We propose a fully operational framework to study the non-equilibrium thermodynamics of a quantum system $S$ that is coupled to a detector $D$ whose state is continuously monitored, allowing to single out individual quantum trajectories of $S$. We focus on detailed fluctuation theorems and characterize the entropy production of the system. We establish fundamental differences with respect to the t…
▽ More
We propose a fully operational framework to study the non-equilibrium thermodynamics of a quantum system $S$ that is coupled to a detector $D$ whose state is continuously monitored, allowing to single out individual quantum trajectories of $S$. We focus on detailed fluctuation theorems and characterize the entropy production of the system. We establish fundamental differences with respect to the thermodynamic of unmonitored, unitarily evolved systems. We consider the paradigmatic example of circuit-QED, where superconducting qubits can be coupled to a continuously monitored resonator and show numerical simulations using state of the art experimental parameters.
△ Less
Submitted 26 October, 2018; v1 submitted 3 April, 2017;
originally announced April 2017.
-
Coherent manipulation of noise-protected superconducting artificial atoms in the Lambda scheme
Authors:
P. G. Di Stefano,
E. Paladino,
T. J. Pope,
G. Falci
Abstract:
We propose a new protocol for thr manipulation of a three-level artificial atom in Lambda ($Λ$) configuration in the absence of a direct pump coupling. It allows faithful, selective and robust population transfer analogous to stimulated Raman adiabatic passage ($Λ$-STIRAP), in highly noise protected superconducting artificial atoms. It combines the use of a two-photon pump pulse with suitable adva…
▽ More
We propose a new protocol for thr manipulation of a three-level artificial atom in Lambda ($Λ$) configuration in the absence of a direct pump coupling. It allows faithful, selective and robust population transfer analogous to stimulated Raman adiabatic passage ($Λ$-STIRAP), in highly noise protected superconducting artificial atoms. It combines the use of a two-photon pump pulse with suitable advanced control, operated by a slow modulation of the phase of the external fields, leveraging on the stability of semiclassical microwave drives. This protocol is a building block for novel tasks in complex quantum architectures. Its demonstration would be a benchmark for the implementation of a class of multilevel advanced control procedures for quantum computation and microwave quantum photonics in systems based on artificial atoms.
△ Less
Submitted 28 May, 2016; v1 submitted 18 September, 2015;
originally announced September 2015.
-
Population transfer in a Lambda system induced by detunings
Authors:
P. G. Di Stefano,
E. Paladino,
A. D'Arrigo,
G. Falci
Abstract:
In this paper we propose a new protocol to achieve coherent population transfer between two states in a three-level atom by using two ac fields. It is based on the physics of Stimulated Raman Adiabatic Passage (STIRAP), but it is implemented with the constraint of a reduced control, namely one of the fields cannot be switched off. A combination of frequency chirps is used with resonant fields, all…
▽ More
In this paper we propose a new protocol to achieve coherent population transfer between two states in a three-level atom by using two ac fields. It is based on the physics of Stimulated Raman Adiabatic Passage (STIRAP), but it is implemented with the constraint of a reduced control, namely one of the fields cannot be switched off. A combination of frequency chirps is used with resonant fields, allowing to achieve approximate destructive interference, despite of the fact that an exact dark state does not exist. This new chirped STIRAP protocol is tailored for applications to artificial atoms, where architectures with several elementary units can be strongly coupled but where the possibility of switching on and off such couplings is often very limited. Demonstration of this protocol would be a benchmark for the implementation of a class of multilevel advanced control procedures for quantum computation and microwave quantum photonics in artificial atoms.
△ Less
Submitted 13 May, 2015;
originally announced May 2015.
-
Design of a Lambda configuration in artificial coherent nanostructures
Authors:
P. G. Di Stefano,
E. Paladino,
A. D'Arrigo,
B. Spagnolo,
G. Falci
Abstract:
The implementation of a three-level Lambda System in artificial atoms would allow to perform advanced control tasks typical of quantum optics in the solid state realm, with photons in the $\mathrm{μm}$/mm range. However hardware constraints put an obstacle since protection from decoherence is often conflicting with efficient coupling to external fields. We address the problem of performing convent…
▽ More
The implementation of a three-level Lambda System in artificial atoms would allow to perform advanced control tasks typical of quantum optics in the solid state realm, with photons in the $\mathrm{μm}$/mm range. However hardware constraints put an obstacle since protection from decoherence is often conflicting with efficient coupling to external fields. We address the problem of performing conventional STImulated Raman Adiabatic Passage (STIRAP) in the presence of low-frequency noise. We propose two strategies to defeat decoherence, based on "optimal symmetry breaking" and dynamical decoupling. We suggest how to apply to the different implementations of superconducting artificial atoms, stressing the key role of non-Markovianity.
△ Less
Submitted 20 February, 2015;
originally announced February 2015.
