Deloitte & Touche Llp

This article examines the transformative impact of quantum computing integration within cloud-based platforms, focusing on the architectural frameworks, security implications, and enterprise applications. Through analysis of current industry developments and technological capabilities, this article explores how quantum-classical hybrid systems are reshaping traditional cloud computing paradigms. It investigates critical aspects including post-quantum cryptography, data processing optimization, and resource virtualization while addressing implementation challenges faced by organizations adopting quantum cloud solutions. It indicates that successful quantum cloud integration requires a strategic approach encompassing infrastructure adaptation, workforce development, and security protocol enhancement. This article contributes to the emerging field of quantum cloud computing by providing a comprehensive framework for enterprise adoption and highlighting the technological prerequisites for effective implementation. It concludes by presenting strategic recommendations for organizations preparing to leverage quantum capabilities within their cloud infrastructure, emphasizing the importance of scalable and secure hybrid architectures

With the constant flow of data from vast sources over the past decades, a plethora of advanced analytical techniques have been developed to extract relevant information from different data types ranging from labeled data, quasi‐labeled data, and data with no labels known a priori. For data with at best quasi‐labels, graphs are a natural representation and have important applications in many industries and scientific disciplines. Specifically, for unlabeled data, anomaly detection on graphs is a method to determine which data points do not posses the latent characteristics that are present in most other data. There have been a variety of classical methods to compute an anomalous score for the individual vertices of a respective graph, such as checking the local topology of a node, random walks, and complex neural networks. Leveraging the structure of the graph, the first quantum algorithm is proposed to calculate the anomaly score of each node by continuously traversing the graph with a uniform starting position for all nodes. The proposed algorithm incorporates well‐known characteristics of quantum walks, and, taking into consideration the noisy intermediate‐scale quantum (NISQ) era and subsequent intermediate‐scale quantum (ISQ) era, an adjustment to the algorithm is provided to mitigate the increasing depth of the circuit. This algorithm is rigorously shown to converge to the expected probability with respect to the initial condition.

We report the results of a field experiment designed to increase honest disclosure of claims at a U.S. state unemployment agency. Individuals filing claims were randomized to a message (‘nudge’) intervention, while an off-the-shelf machine learning algorithm calculated claimants’ risk for committing fraud (underreporting earnings). We study the causal effects of algorithmic targeting on the effectiveness of nudge messages: Without algorithmic targeting, the average treatment effect of the messages was insignificant; in contrast, the use of algorithmic targeting revealed significant heterogeneous treatment effects across claimants. Claimants predicted to behave unethically by the algorithm were more likely to disclose earnings when receiving a message relative to a control condition, with claimants predicted to most likely behave unethically being almost twice as likely to disclose earnings when shown a message. In addition to providing a potential blueprint for targeting more costly interventions, our study offers a novel perspective for the use and efficiency of data science in the public sector without violating citizens’ agency. However, we caution that, while algorithms can enable tailored policy, their ethical use must be ensured at all times.

Accurately predicting individual antidepressant treatment response could expedite the lengthy trial‐and‐error process of finding an effective treatment for major depressive disorder (MDD). We tested and compared machine learning‐based methods that predict individual‐level pharmacotherapeutic treatment response using cortical morphometry from multisite longitudinal cohorts. We conducted an international analysis of pooled data from six sites of the ENIGMA‐MDD consortium (n = 262 MDD patients; age = 36.5 ± 15.3 years; 154 (59%) female; mean response rate = 57%). Treatment response was defined as a ≥ 50% reduction in symptom severity score after 4–12 weeks post‐initiation of antidepressant treatment. Structural MRI was acquired before, or < 14 days after, treatment initiation. The cortex was parcellated using FreeSurfer, from which cortical thickness and surface area were measured. We tested several machine learning pipeline configurations, which varied in (i) the way we presented the cortical data (i.e., average values per region of interest, as a vector containing voxel‐wise cortical thickness and surface area measures, and as cortical thickness and surface area projections), (ii) whether we included clinical data, and the (iii) machine learning model (i.e., gradient boosting, support vector machine, and neural network classifiers) and (iv) cross‐validation methods (i.e., k‐fold and leave‐one‐site‐out) we used. First, we tested if the overall predictive performance of the pipelines was better than chance, with a corrected 10‐fold cross‐validation permutation test. Second, we compared if some machine learning pipeline configurations outperformed others. In an exploratory analysis, we repeated our first analysis in three subpopulations, namely patients (i) from a single site, (ii) with comparable response rates, and (iii) showing the least (first quartile) and the most (fourth quartile) treatment response, which we call the extreme (non‐)responders subpopulation. Finally, we explored the effect of including subcortical volumetric data on model performance. Overall, performance predicting antidepressant treatment response was not significantly better than chance (balanced accuracy = 50.5%; p = 0.66) and did not vary with alternative pipeline configurations. Exploratory analyses revealed that performance across models was only significantly better than chance in the extreme (non‐)responders subpopulation (balanced accuracy = 63.9%, p = 0.001). Including subcortical data did not alter the observed model performance. Cortical structural MRI alone could not reliably predict individual pharmacotherapeutic treatment response in MDD. None of the used machine learning pipeline configurations outperformed the others. In exploratory analyses, we found that predicting response in the extreme (non‐)responders subpopulation was feasible on both cortical data alone and combined with subcortical data, which suggests that specific MDD subpopulations may exhibit response‐related patterns in structural data. Future work may use multimodal data to predict treatment response in MDD.

Human activity and behavior recognition has become part and parcel of remote health monitoring and home automation systems. Our recent works have shown initial success in employing dictionary learning (DL) based solutions for such systems and K-SVD algorithms and its variants have been proposed and successfully utilized in this regard. Motivated by those, our present work further investigates such problems and shows how a novel dictionary learning (DL) based solution using Discriminative K-SVD algorithm can be proposed to address such human activity recognition problems. In order to strengthen the performance and address more challenging human activities, this paper introduces a novel dictionary learning approach, named as Projected Orthogonal Matching Pursuit-Discriminative K-SVD (POMP-DK-SVD), hybridizing characteristic features of POMP and DK-SVD algorithm. The approach, in spirit, proposes to integrate POMP within the sparse coding stage of DK-SVD. The supremacy of the proposed POMP-DK-SVD approach has been firmly established over earlier published K-SVD and modified K-SVD based approaches, by considering two benchmark human activity recognition problems at hand.

Emotional communication is central to strong interpersonal relationships and mental health. For emerging adults, much of their interpersonal communication happens via texting, but we know little about how and with whom their emotional texting occurs. Understanding these patterns may be especially critical for emerging adults experiencing mental health crises (e.g., suicidality). This study examined how the emotional tone of texts (sent and received) varies based on texting partner (i.e., peers, which includes friends and significant others, vs. family) and within-person suicide risk. Participants were 27 emerging adults with prior, non-lethal suicide attempt(s). Linguistic Inquiry Word Count was used to label 75,928 texts. While there was variation in the specific interactions, results overall indicated that participants exchanged texts containing all emotion category words at greater rates with peers (vs. family), and this pattern generally held across within-person suicide risk level. Implications for suicide prevention/intervention efforts are discussed.

Feature selection is a technique in statistical prediction modeling that identifies features in a record with a strong statistical connection to the target variable. Excluding features with a weak statistical connection to the target variable in training not only drops the dimension of the data, which decreases the time complexity of the algorithm, it also decreases noise within the data which assists in avoiding overfitting. In all, feature selection assists in training a robust statistical model that performs well and is stable. Recent advancements in feature selection that leverages quantum annealing (QA) give a scalable technique that aims to maximize the predictive power of the features while minimizing redundancy. As a consequence, it is expected that this algorithm would assist in the bias/variance trade-off yielding better features for training a statistical model. This paper tests this intuition against classical methods by utilizing open-source data sets and evaluates the efficacy of each trained statistical model well-known prediction algorithms. The numerical results display an advantage utilizing the features selected from the algorithm that leveraged QA.