For neurological disorders and diseases, functional and anatomical connectomes of the human brain can be used to better inform targeted interventions and treatment strategies. Functional magnetic resonance imaging (fMRI) is a non-invasive neuroimaging technique that captures spatio-temporal brain function through blood flow over time. FMRI can be used to study the functional connectome through the functional connectivity matrix; that is, Pearson's correlation matrix between time series from the regions of interest of an fMRI image. One approach to analysing functional connectivity is using partial least squares (PLS), a multivariate regression technique designed for high-dimensional predictor data. However, analysing functional connectivity with PLS ignores a key property of the functional connectivity matrix; namely, these matrices are positive definite. To account for this, we introduce a generalisation of PLS to Riemannian manifolds, called R-PLS, and apply it to symmetric positive definite matrices with the affine invariant geometry. We apply R-PLS to two functional imaging datasets: COBRE, which investigates functional differences between schizophrenic patients and healthy controls, and; ABIDE, which compares people with autism spectrum disorder and neurotypical controls. Using the variable importance in the projection statistic on the results of R-PLS, we identify key functional connections in each dataset that are well represented in the literature. Given the generality of R-PLS, this method has potential to open up new avenues for multi-model imaging analysis linking structural and functional connectomics.

Feature transformation aims to reconstruct an effective representation space by mathematically refining the existing features. It serves as a pivotal approach to combat the curse of dimensionality, enhance model generalization, mitigate data sparsity, and extend the applicability of classical models. Existing research predominantly focuses on domain knowledge-based feature engineering or learning latent representations. However, these methods, while insightful, lack full automation and fail to yield a traceable and optimal representation space. An indispensable question arises: Can we concurrently address these limitations when reconstructing a feature space for a machine-learning task? Our initial work took a pioneering step towards this challenge by introducing a novel self-optimizing framework. This framework leverages the power of three cascading reinforced agents to automatically select candidate features and operations for generating improved feature transformation combinations. Despite the impressive strides made, there was room for enhancing its effectiveness and generalization capability. In this extended journal version, we advance our initial work from two distinct yet interconnected perspectives: 1) We propose a refinement of the original framework, which integrates a graph-based state representation method to capture the feature interactions more effectively and develop different Q-learning strategies to alleviate Q-value overestimation further. 2) We utilize a new optimization technique (actor-critic) to train the entire self-optimizing framework in order to accelerate the model convergence and improve the feature transformation performance. Finally, to validate the improved effectiveness and generalization capability of our framework, we perform extensive experiments and conduct comprehensive analyses.

Physicians considering clinical trials for their patients are met with the laborious process of checking many text based eligibility criteria. Large Language Models (LLMs) have shown to perform well for clinical information extraction and clinical reasoning, including medical tests, but not yet in real-world scenarios. This paper investigates the use of InstructGPT to assist physicians in determining eligibility for clinical trials based on a patient's summarised medical profile. Using a prompting strategy combining one-shot, selection-inference and chain-of-thought techniques, we investigate the performance of LLMs on 10 synthetically created patient profiles. Performance is evaluated at four levels: ability to identify screenable eligibility criteria from a trial given a medical profile; ability to classify for each individual criterion whether the patient qualifies; the overall classification whether a patient is eligible for a clinical trial and the percentage of criteria to be screened by physician. We evaluated against 146 clinical trials and a total of 4,135 eligibility criteria. The LLM was able to correctly identify the screenability of 72% (2,994/4,135) of the criteria. Additionally, 72% (341/471) of the screenable criteria were evaluated correctly. The resulting trial level classification as eligible or ineligible resulted in a recall of 0.5. By leveraging LLMs with a physician-in-the-loop, a recall of 1.0 and precision of 0.71 on clinical trial level can be achieved while reducing the amount of criteria to be checked by an estimated 90%. LLMs can be used to assist physicians with pre-screening of patients for clinical trials. By forcing instruction-tuned LLMs to produce chain-of-thought responses, the reasoning can be made transparent to and the decision process becomes amenable by physicians, thereby making such a system feasible for use in real-world scenarios.

Sports betting's recent federal legalisation in the USA coincides with the golden age of machine learning. If bettors can leverage data to reliably predict the probability of an outcome, they can recognise when the bookmaker's odds are in their favour. As sports betting is a multi-billion dollar industry in the USA alone, identifying such opportunities could be extremely lucrative. Many researchers have applied machine learning to the sports outcome prediction problem, generally using accuracy to evaluate the performance of predictive models. We hypothesise that for the sports betting problem, model calibration is more important than accuracy. To test this hypothesis, we train models on NBA data over several seasons and run betting experiments on a single season, using published odds. We show that optimising the predictive model for calibration leads to greater returns than optimising for accuracy, on average (return on investment of $+34.69\%$ versus $-35.17\%$) and in the best case ($+36.93\%$ versus $+5.56\%$). These findings suggest that for sports betting (or any probabilistic decision-making problem), calibration is a more important metric than accuracy. Sports bettors who wish to increase profits should therefore optimise their predictive model for calibration.

There is a longstanding interest in capturing the error behaviour of object detectors by finding images where their performance is likely to be unsatisfactory. In real-world applications such as autonomous driving, it is also crucial to characterise potential failures beyond simple requirements of detection performance. For example, a missed detection of a pedestrian close to an ego vehicle will generally require closer inspection than a missed detection of a car in the distance. The problem of predicting such potential failures at test time has largely been overlooked in the literature and conventional approaches based on detection uncertainty fall short in that they are agnostic to such fine-grained characterisation of errors. In this work, we propose to reformulate the problem of finding "hard" images as a query-based hard image retrieval task, where queries are specific definitions of "hardness", and offer a simple and intuitive method that can solve this task for a large family of queries. Our method is entirely post-hoc, does not require ground-truth annotations, is independent of the choice of a detector, and relies on an efficient Monte Carlo estimation that uses a simple stochastic model in place of the ground-truth. We show experimentally that it can be applied successfully to a wide variety of queries for which it can reliably identify hard images for a given detector without any labelled data. We provide results on ranking and classification tasks using the widely used RetinaNet, Faster-RCNN, Mask-RCNN, and Cascade Mask-RCNN object detectors. The code for this project is available at https://github.com/fiveai/hardest.

As machine learning (ML) based systems are adopted in domains such as law enforcement, criminal justice, finance, hiring and admissions, ensuring the fairness of ML aided decision-making is becoming increasingly important. In this paper, we focus on the problem of fair classification, and introduce a novel min-max F-divergence regularization framework for learning fair classification models while preserving high accuracy. Our framework consists of two trainable networks, namely, a classifier network and a bias/fairness estimator network, where the fairness is measured using the statistical notion of F-divergence. We show that F-divergence measures possess convexity and differentiability properties, and their variational representation make them widely applicable in practical gradient based training methods. The proposed framework can be readily adapted to multiple sensitive attributes and for high dimensional datasets. We study the F-divergence based training paradigm for two types of group fairness constraints, namely, demographic parity and equalized odds. We present a comprehensive set of experiments for several real-world data sets arising in multiple domains (including COMPAS, Law Admissions, Adult Income, and CelebA datasets). To quantify the fairness-accuracy tradeoff, we introduce the notion of fairness-accuracy receiver operating characteristic (FA-ROC) and a corresponding \textit{low-bias} FA-ROC, which we argue is an appropriate measure to evaluate different classifiers. In comparison to several existing approaches for learning fair classifiers (including pre-processing, post-processing and other regularization methods), we show that the proposed F-divergence based framework achieves state-of-the-art performance with respect to the trade-off between accuracy and fairness.

Despite the popularity and success of deep learning, there is limited understanding of when, how, and why neural networks generalize to unseen examples. Since learning can be seen as extracting information from data, we formally study information captured by neural networks during training. Specifically, we start with viewing learning in presence of noisy labels from an information-theoretic perspective and derive a learning algorithm that limits label noise information in weights. We then define a notion of unique information that an individual sample provides to the training of a deep network, shedding some light on the behavior of neural networks on examples that are atypical, ambiguous, or belong to underrepresented subpopulations. We relate example informativeness to generalization by deriving nonvacuous generalization gap bounds. Finally, by studying knowledge distillation, we highlight the important role of data and label complexity in generalization. Overall, our findings contribute to a deeper understanding of the mechanisms underlying neural network generalization.

Recommender Systems (RS) currently represent a fundamental tool in online services, especially with the advent of Online Social Networks (OSN). In this case, users generate huge amounts of contents and they can be quickly overloaded by useless information. At the same time, social media represent an important source of information to characterize contents and users' interests. RS can exploit this information to further personalize suggestions and improve the recommendation process. In this paper we present a survey of Recommender Systems designed and implemented for Online and Mobile Social Networks, highlighting how the use of social context information improves the recommendation task, and how standard algorithms must be enhanced and optimized to run in a fully distributed environment, as opportunistic networks. We describe advantages and drawbacks of these systems in terms of algorithms, target domains, evaluation metrics and performance evaluations. Eventually, we present some open research challenges in this area.

Recent results of machine learning for automatic vulnerability detection have been very promising indeed: Given only the source code of a function $f$, models trained by machine learning techniques can decide if $f$ contains a security flaw with up to 70% accuracy. But how do we know that these results are general and not specific to the datasets? To study this question, researchers proposed to amplify the testing set by injecting semantic preserving changes and found that the model's accuracy significantly drops. In other words, the model uses some unrelated features during classification. In order to increase the robustness of the model, researchers proposed to train on amplified training data, and indeed model accuracy increased to previous levels. In this paper, we replicate and continue this investigation, and provide an actionable model benchmarking methodology to help researchers better evaluate advances in machine learning for vulnerability detection. Specifically, we propose (i) a cross validation algorithm, where a semantic preserving transformation is applied during the amplification of either the training set or the testing set, and (ii) the amplification of the testing set with code snippets where the vulnerabilities are fixed. Using 11 transformations, 3 ML techniques, and 2 datasets, we find that the improved robustness only applies to the specific transformations used during training data amplification. In other words, the robustified models still rely on unrelated features for predicting the vulnerabilities in the testing data. Additionally, we find that the trained models are unable to generalize to the modified setting which requires to distinguish vulnerable functions from their patches.

This paper investigates the impact of LiDAR configuration shifts on the performance of 3D LiDAR point cloud semantic segmentation models, a topic not extensively studied before. We explore the effect of using different LiDAR channels when training and testing a 3D LiDAR point cloud semantic segmentation model, utilizing Cylinder3D for the experiments. A Cylinder3D model is trained and tested on simulated 3D LiDAR point cloud datasets created using the Mississippi State University Autonomous Vehicle Simulator (MAVS) and 32, 64 channel 3D LiDAR point clouds of the RELLIS-3D dataset collected in a real-world off-road environment. Our experimental results demonstrate that sensor and spatial domain shifts significantly impact the performance of LiDAR-based semantic segmentation models. In the absence of spatial domain changes between training and testing, models trained and tested on the same sensor type generally exhibited better performance. Moreover, higher-resolution sensors showed improved performance compared to those with lower-resolution ones. However, results varied when spatial domain changes were present. In some cases, the advantage of a sensor's higher resolution led to better performance both with and without sensor domain shifts. In other instances, the higher resolution resulted in overfitting within a specific domain, causing a lack of generalization capability and decreased performance when tested on data with different sensor configurations.