Model selection for a given target task can be costly, as it may entail extensive annotation of the quality of outputs of different models. We introduce DiffUse, an efficient method to make an informed decision between candidate text generation models. DiffUse reduces the required amount of preference annotations, thus saving valuable time and resources in performing evaluation. DiffUse intelligently selects instances by clustering embeddings that represent the semantic differences between model outputs. Thus, it is able to identify a subset of examples that are more informative for preference decisions. Our method is model-agnostic, and can be applied to any text generation model. Moreover, we propose a practical iterative approach for dynamically determining how many instances to annotate. In a series of experiments over hundreds of model pairs, we demonstrate that DiffUse can dramatically reduce the required number of annotations -- by up to 75% -- while maintaining high evaluation reliability.

Early detection of network intrusions and cyber threats is one of the main pillars of cybersecurity. One of the most effective approaches for this purpose is to analyze network traffic with the help of artificial intelligence algorithms, with the aim of detecting the possible presence of an attacker by distinguishing it from a legitimate user. This is commonly done by collecting the traffic exchanged between terminals in a network and analyzing it on a per-packet or per-connection basis. In this paper, we propose instead to perform pre-processing of network traffic under analysis with the aim of extracting some new metrics on which we can perform more efficient detection and overcome some limitations of classical approaches. These new metrics are based on graph theory, and consider the network as a whole, rather than focusing on individual packets or connections. Our approach is validated through experiments performed on publicly available data sets, from which it results that it can not only overcome some of the limitations of classical approaches, but also achieve a better detection capability of cyber threats.

In modern machine learning, models can often fit training data in numerous ways, some of which perform well on unseen (test) data, while others do not. Remarkably, in such cases gradient descent frequently exhibits an implicit bias that leads to excellent performance on unseen data. This implicit bias was extensively studied in supervised learning, but is far less understood in optimal control (reinforcement learning). There, learning a controller applied to a system via gradient descent is known as policy gradient, and a question of prime importance is the extent to which a learned controller extrapolates to unseen initial states. This paper theoretically studies the implicit bias of policy gradient in terms of extrapolation to unseen initial states. Focusing on the fundamental Linear Quadratic Regulator (LQR) problem, we establish that the extent of extrapolation depends on the degree of exploration induced by the system when commencing from initial states included in training. Experiments corroborate our theory, and demonstrate its conclusions on problems beyond LQR, where systems are non-linear and controllers are neural networks. We hypothesize that real-world optimal control may be greatly improved by developing methods for informed selection of initial states to train on.

In this paper, we propose a novel approach to Bayesian Experimental Design (BED) for non-exchangeable data that formulates it as risk-sensitive policy optimization. We develop the Inside-Out SMC^2 algorithm that uses a nested sequential Monte Carlo (SMC) estimator of the expected information gain and embeds it into a particle Markov chain Monte Carlo (pMCMC) framework to perform gradient-based policy optimization. This is in contrast to recent approaches that rely on biased estimators of the expected information gain (EIG) to amortize the cost of experiments by learning a design policy in advance. Numerical validation on a set of dynamical systems showcases the efficacy of our method in comparison to other state-of-the-art strategies.

In the clinical treatment of mood disorders, the complex behavioral symptoms presented by patients and variability of patient response to particular medication classes can create difficulties in providing fast and reliable treatment when standard diagnostic and prescription methods are used. Increasingly, the incorporation of physiological information such as neuroimaging scans and derivatives into the clinical process promises to alleviate some of the uncertainty surrounding this process. Particularly, if neural features can help to identify patients who may not respond to standard courses of anti-depressants or mood stabilizers, clinicians may elect to avoid lengthy and side-effect-laden treatments and seek out a different, more effective course that might otherwise not have been under consideration. Previously, approaches for the derivation of relevant neuroimaging features work at only one scale in the data, potentially limiting the depth of information available for clinical decision support. In this work, we show that the utilization of multi spatial scale neuroimaging features - particularly resting state functional networks and functional network connectivity measures - provide a rich and robust basis for the identification of relevant medication class and non-responders in the treatment of mood disorders. We demonstrate that the generated features, along with a novel approach for fast and automated feature selection, can support high accuracy rates in the identification of medication class and non-responders as well as the identification of novel, multi-scale biomarkers.

In the field of domain generalization, the task of constructing a predictive model capable of generalizing to a target domain without access to target data remains challenging. This problem becomes further complicated when considering evolving dynamics between domains. While various approaches have been proposed to address this issue, a comprehensive understanding of the underlying generalization theory is still lacking. In this study, we contribute novel theoretic results that aligning conditional distribution leads to the reduction of generalization bounds. Our analysis serves as a key motivation for solving the Temporal Domain Generalization (TDG) problem through the application of Koopman Neural Operators, resulting in Temporal Koopman Networks (TKNets). By employing Koopman Operators, we effectively address the time-evolving distributions encountered in TDG using the principles of Koopman theory, where measurement functions are sought to establish linear transition relations between evolving domains. Through empirical evaluations conducted on synthetic and real-world datasets, we validate the effectiveness of our proposed approach.

Consider a multi-class labelling problem, where the labels can take values in $[k]$, and a predictor predicts a distribution over the labels. In this work, we study the following foundational question: Are there notions of multi-class calibration that give strong guarantees of meaningful predictions and can be achieved in time and sample complexities polynomial in $k$? Prior notions of calibration exhibit a tradeoff between computational efficiency and expressivity: they either suffer from having sample complexity exponential in $k$, or needing to solve computationally intractable problems, or give rather weak guarantees. Our main contribution is a notion of calibration that achieves all these desiderata: we formulate a robust notion of projected smooth calibration for multi-class predictions, and give new recalibration algorithms for efficiently calibrating predictors under this definition with complexity polynomial in $k$. Projected smooth calibration gives strong guarantees for all downstream decision makers who want to use the predictor for binary classification problems of the form: does the label belong to a subset $T \subseteq [k]$: e.g. is this an image of an animal? It ensures that the probabilities predicted by summing the probabilities assigned to labels in $T$ are close to some perfectly calibrated binary predictor for that task. We also show that natural strengthenings of our definition are computationally hard to achieve: they run into information theoretic barriers or computational intractability. Underlying both our upper and lower bounds is a tight connection that we prove between multi-class calibration and the well-studied problem of agnostic learning in the (standard) binary prediction setting.

Scientific modeling applications often require estimating a distribution of parameters consistent with a dataset of observations - an inference task also known as source distribution estimation. This problem can be ill-posed, however, since many different source distributions might produce the same distribution of data-consistent simulations. To make a principled choice among many equally valid sources, we propose an approach which targets the maximum entropy distribution, i.e., prioritizes retaining as much uncertainty as possible. Our method is purely sample-based - leveraging the Sliced-Wasserstein distance to measure the discrepancy between the dataset and simulations - and thus suitable for simulators with intractable likelihoods. We benchmark our method on several tasks, and show that it can recover source distributions with substantially higher entropy without sacrificing the fidelity of the simulations. Finally, to demonstrate the utility of our approach, we infer source distributions for parameters of the Hodgkin-Huxley neuron model from experimental datasets with thousands of measurements. In summary, we propose a principled framework for inferring unique source distributions of scientific simulator parameters while retaining as much uncertainty as possible.

Large-scale machine learning problems make the cost of hyperparameter tuning ever more prohibitive. This creates a need for algorithms that can tune themselves on-the-fly. We formalize the notion of "tuning-free" algorithms that can match the performance of optimally-tuned optimization algorithms up to polylogarithmic factors given only loose hints on the relevant problem parameters. We consider in particular algorithms that can match optimally-tuned Stochastic Gradient Descent (SGD). When the domain of optimization is bounded, we show tuning-free matching of SGD is possible and achieved by several existing algorithms. We prove that for the task of minimizing a convex and smooth or Lipschitz function over an unbounded domain, tuning-free optimization is impossible. We discuss conditions under which tuning-free optimization is possible even over unbounded domains. In particular, we show that the recently proposed DoG and DoWG algorithms are tuning-free when the noise distribution is sufficiently well-behaved. For the task of finding a stationary point of a smooth and potentially nonconvex function, we give a variant of SGD that matches the best-known high-probability convergence rate for tuned SGD at only an additional polylogarithmic cost. However, we also give an impossibility result that shows no algorithm can hope to match the optimal expected convergence rate for tuned SGD with high probability.

Binary classification tasks are prevalent in learning algorithms, as diverse scenarios require binary decisions. Examples include predicting default risk or accident occurrence in insurance or finance as well as disease likelihood in healthcare. To improve reliability, particularly in sensitive decision-making contexts, a classifier must possess strong discriminatory capabilities. Typically, classifiers are trained to optimize goodness-of-fit criteria, often based on the accuracy of class predictions. However, goodness-of-fit criteria, such as accuracy or AUC, do not consider the varying confidence levels assigned by the algorithm to each prediction. If the sole objective is effective class prediction, then the classifier fulfills its purpose. Nevertheless, there are instances where interest extends beyond the predicted class to the associated likelihood. This occurs when predicting loan repayment defaults (Liu et al., 2021) or accident incidences, as risk transfer pricing is usually tied directly to event probabilities. In such cases, the model-predicted scores of classifiers are often interpreted as event probabilities.