arXiv subjects

Arnoldo Frigessi

Publications and source records attributed to Arnoldo Frigessi.

At least 19 recordsLinked to original sources

arXiv2026

Stepwise Variational Inference with Vine Copulas

We propose stepwise variational inference (VI) with vine copulas: a universal VI procedure that combines vine copulas with a novel stepwise estimation procedure of the variational parameters. Vine copulas consist of a nested sequence of trees built from copulas, where more complex latent dependence can be modeled with increasing number of trees. We propose to estimate the vine copula approximate posterior in a stepwise fashion, tree by tree along the vine structure. Further, we show that the usual backward Kullback-Leibler divergence cannot recover the correct parameters in the vine copula model, thus the evidence lower bound is defined based on the R\'enyi divergence. Finally, an intuitive stopping criterion for adding further trees to the vine eliminates the need to pre-define a complexity parameter of the variational distribution, as required for most other approaches. Thus, our method interpolates between mean-field VI (MFVI) and full latent dependence. In many applications, in particular sparse Gaussian processes, our method is parsimonious with parameters, while outperforming MFVI.

stat.ML↗

arXiv2026

Beyond Accuracy: Decomposing the Reasoning Efficiency of LLMs

As reasoning LLMs increasingly trade tokens for accuracy through deliberation, search, and self-correction, a single accuracy score can no longer tell whether those tokens buy useful reasoning, recovery from hard instances, or unnecessary verbosity. We introduce a trace-optional evaluation protocol that exactly decomposes token efficiency using three observables available even for closed models: completion rate, conditional correctness given completion, and generated length. When instance-level workload metadata is available, we further normalize generated length by declared task-implied work and separate mean verbalization overhead from workload-dependent scaling. When such metadata is absent, we define an auditable solver-derived workload scale and evaluate its stability under leave-self-out, leave-top-k, and held-out-reference-pool perturbations. We evaluate 14 shared open-weight models on CogniLoad, GSM8K, ProofWriter, and ZebraLogic. We further evaluate 11 additional models on CogniLoad, enabling a fine-grained analysis of reasoning-task difficulty factors: task length, intrinsic difficulty, and distractor density. Efficiency and overhead rankings remain stable across all benchmark pairs, more robustly than accuracy rankings, while the decomposition separates logic-limited, context-limited (truncation-driven), and verbosity-limited failure modes that look identical under accuracy-per-token. We release an evaluation artifact and reporting template, which elaborates on why an LLM is inefficient at reasoning.

cs.CL↗

arXiv2025

CogniLoad: A Synthetic Natural Language Reasoning Benchmark With Tunable Length, Intrinsic Difficulty, and Distractor Density

Current benchmarks for long-context reasoning in Large Language Models (LLMs) often blur critical factors like intrinsic task complexity, distractor interference, and task length. To enable more precise failure analysis, we introduce CogniLoad, a novel synthetic benchmark grounded in Cognitive Load Theory (CLT). CogniLoad generates natural-language logic puzzles with independently tunable parameters that reflect CLT's core dimensions: intrinsic difficulty ($d$) controls intrinsic load; distractor-to-signal ratio ($ρ$) regulates extraneous load; and task length ($N$) serves as an operational proxy for conditions demanding germane load. Evaluating 22 SotA reasoning LLMs, CogniLoad reveals distinct performance sensitivities, identifying task length as a dominant constraint and uncovering varied tolerances to intrinsic complexity and U-shaped responses to distractor ratios. By offering systematic, factorial control over these cognitive load dimensions, CogniLoad provides a reproducible, scalable, and diagnostically rich tool for dissecting LLM reasoning limitations and guiding future model development.

cs.CL↗

arXiv2025

TVineSynth: A Truncated C-Vine Copula Generator of Synthetic Tabular Data to Balance Privacy and Utility

We propose TVineSynth, a vine copula based synthetic tabular data generator, which is designed to balance privacy and utility, using the vine tree structure and its truncation to do the trade-off. Contrary to synthetic data generators that achieve DP by globally adding noise, TVineSynth performs a controlled approximation of the estimated data generating distribution, so that it does not suffer from poor utility of the resulting synthetic data for downstream prediction tasks. TVineSynth introduces a targeted bias into the vine copula model that, combined with the specific tree structure of the vine, causes the model to zero out privacy-leaking dependencies while relying on those that are beneficial for utility. Privacy is here measured with membership (MIA) and attribute inference attacks (AIA). Further, we theoretically justify how the construction of TVineSynth ensures AIA privacy under a natural privacy measure for continuous sensitive attributes. When compared to competitor models, with and without DP, on simulated and on real-world data, TVineSynth achieves a superior privacy-utility balance.

cs.LG↗

arXiv2025

BoRA: Bayesian Hierarchical Low-Rank Adaption for Multi-Task Large Language Models

This paper introduces Bayesian Hierarchical Low-Rank Adaption (BoRA), a novel method for finetuning multi-task Large Language Models (LLMs). Current finetuning approaches, such as Low-Rank Adaption (LoRA), perform exeptionally well in reducing training parameters and memory usage but face limitations when applied to multiple similar tasks. Practitioners usually have to choose between training separate models for each task or a single model for all tasks, both of which come with trade-offs in specialization and data utilization. BoRA addresses these trade-offs by leveraging a Bayesian hierarchical model that allows tasks to share information through global hierarchical priors. This enables tasks with limited data to benefit from the overall structure derived from related tasks while allowing tasks with more data to specialize. Our experimental results show that BoRA outperforms both individual and unified model approaches, achieving lower perplexity and better generalization across tasks. This method provides a scalable and efficient solution for multi-task LLM finetuning, with significant practical implications for diverse applications.

cs.LG↗

arXiv2024

Prediction of cancer dynamics under treatment using Bayesian neural networks: A simulated study

Predicting cancer dynamics under treatment is challenging due to high inter-patient heterogeneity, lack of predictive biomarkers, and sparse and noisy longitudinal data. Mathematical models can summarize cancer dynamics by a few interpretable parameters per patient. Machine learning methods can then be trained to predict the model parameters from baseline covariates, but do not account for uncertainty in the parameter estimates. Instead, hierarchical Bayesian modeling can model the relationship between baseline covariates to longitudinal measurements via mechanistic parameters while accounting for uncertainty in every part of the model. The mapping from baseline covariates to model parameters can be modeled in several ways. A linear mapping simplifies inference but fails to capture nonlinear covariate effects and scale poorly for interaction modeling when the number of covariates is large. In contrast, Bayesian neural networks can potentially discover interactions between covariates automatically, but at a substantial cost in computational complexity. In this work, we develop a hierarchical Bayesian model of subpopulation dynamics that uses baseline covariate information to predict cancer dynamics under treatment, inspired by cancer dynamics in multiple myeloma (MM), where serum M protein is a well-known proxy of tumor burden. As a working example, we apply the model to a simulated dataset and compare its ability to predict M protein trajectories to a model with linear covariate effects. Our results show that the Bayesian neural network covariate effect model predicts cancer dynamics more accurately than a linear covariate effect model when covariate interactions are present. The framework can also be applied to other types of cancer or other time series prediction problems that can be described with a parametric model.

q-bio.QM↗

arXiv2024

Rank-based Bayesian clustering via covariate-informed Mallows mixtures

Data in the form of rankings, ratings, pair comparisons or clicks are frequently collected in diverse fields, from marketing to politics, to understand assessors' individual preferences. Combining such preference data with features associated with the assessors can lead to a better understanding of the assessors' behaviors and choices. The Mallows model is a popular model for rankings, as it flexibly adapts to different types of preference data, and the previously proposed Bayesian Mallows Model (BMM) offers a computationally efficient framework for Bayesian inference, also allowing capturing the users' heterogeneity via a finite mixture. We develop a Bayesian Mallows-based finite mixture model that performs clustering while also accounting for assessor-related features, called the Bayesian Mallows model with covariates (BMMx). BMMx is based on a similarity function that a priori favours the aggregation of assessors into a cluster when their covariates are similar, using the Product Partition models (PPMx) proposal. We present two approaches to measure the covariate similarity: one based on a novel deterministic function measuring the covariates' goodness-of-fit to the cluster, and one based on an augmented model as in PPMx. We investigate the performance of BMMx in both simulation experiments and real-data examples, showing the method's potential for advancing the understanding of assessor preferences and behaviors in different applications.

stat.ME↗

arXiv2023

Modeling of mouse experiments suggests that optimal anti-hormonal treatment for breast cancer is diet-dependent

Estrogen receptor positive breast cancer is frequently treated with anti-hormonal treatment such as aromatase inhibitors (AI). Interestingly, a high body mass index has been shown to have a negative impact on AI efficacy, most likely due to disturbances in steroid metabolism and adipokine production. Here, we propose a mathematical model based on a system of ordinary differential equations to investigate the effect of high-fat diet on tumor growth. We inform the model with data from mouse experiments, where the animals are fed with high-fat or control (normal) diet. By incorporating AI treatment with drug resistance into the model and by solving optimal control problems we found differential responses for control and high-fat diet. To the best of our knowledge, this is the first attempt to model optimal anti-hormonal treatment for breast cancer in the presence of drug resistance. Our results underline the importance of considering high-fat diet and obesity as factors influencing clinical outcomes during anti-hormonal therapies in breast cancer patients.

math.OC↗

arXiv2022

ABC of the Future

Approximate Bayesian computation (ABC) has advanced in two decades from a seminal idea to a practically applicable inference tool for simulator-based statistical models, which are becoming increasingly popular in many research domains. The computational feasibility of ABC for practical applications has been recently boosted by adopting techniques from machine learning to build surrogate models for the approximate likelihood or posterior and by the introduction of a general-purpose software platform with several advanced features, including automated parallelization. Here we demonstrate the strengths of the advances in ABC by going beyond the typical benchmark examples and considering real applications in astronomy, infectious disease epidemiology, personalised cancer therapy and financial prediction. We anticipate that the emerging success of ABC in producing actual added value and quantitative insights in the real world will continue to inspire a plethora of further applications across different fields of science, social science and technology.

stat.AP↗

arXiv2022

Effects of Non-Compulsory and Mandatory COVID-19 Interventions on Travel Distance and Time Away from Home: The Case of Norway in 2021

Background: Due to the societal, economic, and health costs of COVID-19 non-pharmaceutical interventions (NPIs), it is important to assess their effects. Human mobility serves as a surrogate for human contacts and compliance to NPIs. In Nordic countries, NPIs have mostly been advised and sometimes made mandatory. It is unclear if making NPIs mandatory further reduced mobility. Aim: We investigated the effect of non-compulsory and follow-up mandatory measures in major cities and rural regions on human mobility in Norway. We identified NPI categories that most affected mobility. Methods: We used mobile phone mobility data from the largest Norwegian operator. We analysed non-compulsory and mandatory measures with before-after and synthetic difference-in-differences approaches. By regression, we investigated the impact of different NPIs on mobility. Results: Nationally and in less populated regions, follow-up mandatory measures further decreased time, but not distance travelled. In urban areas, however, follow-up mandates also decreased distance, and the effect exceeded that of initial non-compulsory measures. Stricter metre rules, gyms closing and reopening, restrictions on guests in homes, and face mask recommendations most impacted distance travelled. Time travelled was most affected by gyms closing and restaurants and shops reopening. Conclusion: Overall, non-compulsory measures appeared to decrease distance travelled from home, while mandates further decreased this metric in urban areas. Time travelled is reduced more by mandates than by non-compulsory measures for all regions and interventions. Stricter distancing and restricted number of guests were associated with decreases in mobility.

physics.soc-ph↗

arXiv2022

Pseudo-Mallows for Efficient Probabilistic Preference Learning

We propose the Pseudo-Mallows distribution over the set of all permutations of $n$ items, to approximate the posterior distribution with a Mallows likelihood. The Mallows model has been proven to be useful for recommender systems where it can be used to learn personal preferences from highly incomplete data provided by the users. Inference based on MCMC is however slow, preventing its use in real time applications. The Pseudo-Mallows distribution is a product of univariate discrete Mallows-like distributions, constrained to remain in the space of permutations. The quality of the approximation depends on the order of the $n$ items used to determine the factorization sequence. In a variational setting, we optimise the variational order parameter by minimising a marginalized KL-divergence. We propose an approximate algorithm for this discrete optimization, and conjecture a certain form of the optimal variational order that depends on the data. Empirical evidence and some theory support our conjecture. Sampling from the Pseudo-Mallows distribution allows fast preference learning, compared to alternative MCMC based options, when the data exists in form of partial rankings of the items or of clicking on some items. Through simulations and a real life data case study, we demonstrate that the Pseudo-Mallows model learns personal preferences very well and makes recommendations much more efficiently, while maintaining similar accuracy compared to the exact Bayesian Mallows model.

stat.ME↗

arXiv2022

A sequential Monte Carlo approach to estimate a time varying reproduction number in infectious disease models: the Covid-19 case

During the first months, the Covid-19 pandemic has required most countries to implement complex sequences of non-pharmaceutical interventions, with the aim of controlling the transmission of the virus in the population. To be able to take rapid decisions, a detailed understanding of the current situation is necessary. Estimates of time-varying, instantaneous reproduction numbers represent a way to quantify the viral transmission in real time. They are often defined through a mathematical compartmental model of the epidemic, like a stochastic SEIR model, whose parameters must be estimated from multiple time series of epidemiological data. Because of very high dimensional parameter spaces (partly due to the stochasticity in the spread models) and incomplete and delayed data, inference is very challenging. We propose a state space formalisation of the model and a sequential Monte Carlo approach which allow to estimate a daily-varying reproduction number for the Covid-19 epidemic in Norway with sufficient precision, on the basis of daily hospitalisation and positive test incidences. The method is in regular use in Norway and is a powerful instrument for epidemic monitoring and management.

q-bio.PE↗

arXiv2021

FINN.no Slates Dataset: A new Sequential Dataset Logging Interactions, allViewed Items and Click Responses/No-Click for Recommender Systems Research

We present a novel recommender systems dataset that records the sequential interactions between users and an online marketplace. The users are sequentially presented with both recommendations and search results in the form of ranked lists of items, called slates, from the marketplace. The dataset includes the presented slates at each round, whether the user clicked on any of these items and which item the user clicked on. Although the usage of exposure data in recommender systems is growing, to our knowledge there is no open large-scale recommender systems dataset that includes the slates of items presented to the users at each interaction. As a result, most articles on recommender systems do not utilize this exposure information. Instead, the proposed models only depend on the user's click responses, and assume that the user is exposed to all the items in the item universe at each step, often called uniform candidate sampling. This is an incomplete assumption, as it takes into account items the user might not have been exposed to. This way items might be incorrectly considered as not of interest to the user. Taking into account the actually shown slates allows the models to use a more natural likelihood, based on the click probability given the exposure set of items, as is prevalent in the bandit and reinforcement learning literature. \cite{Eide2021DynamicSampling} shows that likelihoods based on uniform candidate sampling (and similar assumptions) are implicitly assuming that the platform only shows the most relevant items to the user. This causes the recommender system to implicitly reinforce feedback loops and to be biased towards previously exposed items to the user.

cs.IR↗

arXiv2021

A scalable solver for a stochastic, hybrid cellular automaton model of personalized breast cancer therapy

Mathematical modeling and simulation is a promising approach to personalized cancer medicine. Yet, the complexity, heterogeneity and multi-scale nature of cancer pose significant computational challenges. Coupling discrete cell-based models with continuous models using hybrid cellular automata is a powerful approach for mimicking biological complexity and describing the dynamical exchange of information across different scales. However, when clinically relevant cancer portions are taken into account, such models become computationally very expensive. While efficient parallelization techniques for continuous models exist, their coupling with discrete models, particularly cellular automata, necessitates more elaborate solutions. Building upon FEniCS, a popular and powerful scientific computing platform for solving partial differential equations, we developed parallel algorithms to link stochastic cellular automata with differential equations ( https://bitbucket.org/HTasken/cansim ). The algorithms minimize the communication between processes that share cellular automata neighborhood values while also allowing for reproducibility during stochastic updates. We demonstrated the potential of our solution on a complex hybrid cellular automaton model of breast cancer treated with combination chemotherapy. On a single-core processor, we obtained nearly linear scaling with an increasing problem size, whereas weak parallel scaling showed moderate growth in solving time relative to increase in problem size. Finally we applied the algorithm to a problem that is 500 times larger than previous work, allowing us to run personalized therapy simulations based on heterogeneous cell density and tumor perfusion conditions estimated from magnetic resonance imaging data on an unprecedented scale.

q-bio.QM↗

arXiv2021

Dynamic Slate Recommendation with Gated Recurrent Units and Thompson Sampling

We consider the problem of recommending relevant content to users of an internet platform in the form of lists of items, called slates. We introduce a variational Bayesian Recurrent Neural Net recommender system that acts on time series of interactions between the internet platform and the user, and which scales to real world industrial situations. The recommender system is tested both online on real users, and on an offline dataset collected from a Norwegian web-based marketplace, FINN.no, that is made public for research. This is one of the first publicly available datasets which includes all the slates that are presented to users as well as which items (if any) in the slates were clicked on. Such a data set allows us to move beyond the common assumption that implicitly assumes that users are considering all possible items at each interaction. Instead we build our likelihood using the items that are actually in the slate, and evaluate the strengths and weaknesses of both approaches theoretically and in experiments. We also introduce a hierarchical prior for the item parameters based on group memberships. Both item parameters and user preferences are learned probabilistically. Furthermore, we combine our model with bandit strategies to ensure learning, and introduce `in-slate Thompson Sampling' which makes use of the slates to maximise explorative opportunities. We show experimentally that explorative recommender strategies perform on par or above their greedy counterparts. Even without making use of exploration to learn more effectively, click rates increase simply because of improved diversity in the recommended slates.

stat.ML↗

arXiv2020

Towards personalized computer simulations of breast cancer treatment

Cancer pathology is unique to a given individual, and developing personalized diagnostic and treatment protocols are a primary concern. Mathematical modeling and simulation is a promising approach to personalized cancer medicine. Yet, the complexity, heterogeneity and multiscale nature of cancer present severe challenges. One of the major barriers to use mathematical models to predict the outcome of therapeutic regimens in a particular patient lies in their initialization and parameterization in order to reflect individual cancer characteristics accurately. Here we present a study where we used multitype measurements acquired routinely on a single breast tumor, including histopathology, magnetic resonance imaging (MRI), and molecular profiling, to personalize a multiscale hybrid cellular automaton model of breast cancer treated with chemotherapeutic and antiangiogenic agents. We model drug pharmacokinetics and pharmacodynamics at the tumor tissue level but with cellular and subcellular resolution. We simulate those spatio-temporal dynamics in 2D cross-sections of tumor portions over 12-week therapy regimes, resulting in complex and computationally intensive simulations. For such computationally demanding systems, formal statistical inference methods to estimate individual parameters from data have not been feasible in practice to until most recently, after the emergence of machine learning techniques applied to likelihood-free inference methods. Here we use the inference advances provided by Bayesian optimization to fit our model to simulated data of individual patients. In this way, we investigate if some key parameters can be estimated from a series of measurements of cell density in the tumor tissue, as well as how often the measurements need to be taken to allow reliable predictions.

q-bio.TO↗

arXiv2019

A Bayesian approach to study synergistic interaction effects in in-vitro drug combination experiments

In cancer translational research, increasing effort is devoted to the study of the combined effect of two drugs when they are administered simultaneously. In this paper, we introduce a new approach to estimate the part of the effect of the two drugs due to the interaction of the compounds, i.e. which is due to synergistic or antagonistic effects of the two drugs, compared to a reference value representing the condition when the combined compounds do not interact, called zero-interaction. We describe an in-vitro cell viability experiment as a random experiment, by interpreting cell viability as the probability of a cell in the experiment to be viable after treatment, and including information related to different exposure conditions. We propose a flexible Bayesian spline regression framework for modelling the viability surface of two drugs combined as a function of the concentrations. Since the proposed approach is based on a statistical model, it allows to include replicates of the experiments, to evaluate the uncertainty of the estimates, and to perform prediction. We test the model fit and prediction performance on a simulation study, and on an ovarian cancer cell dataset. Posterior estimates of the zero-interaction level and of the synergy term, obtained via adaptive MCMC algorithms, are used to compute interpretable measures of efficacy of the combined experiment, including relative volume under the surface (rVUS) measures to summarise the zero-interaction and synergy terms and a bi-variate alternative to the well-known EC50 measure.

stat.AP↗

arXiv2019

Diverse personalized recommendations with uncertainty from implicit preference data with the Bayesian Mallows Model

Clicking data, which exists in abundance and contains objective user preference information, is widely used to produce personalized recommendations in web-based applications. Current popular recommendation algorithms, typically based on matrix factorizations, often have high accuracy and achieve good clickthrough rates. However, diversity of the recommended items, which can greatly enhance user experiences, is often overlooked. Moreover, most algorithms do not produce interpretable uncertainty quantifications of the recommendations. In this work, we propose the Bayesian Mallows for Clicking Data (BMCD) method, which augments clicking data into compatible full ranking vectors by enforcing all the clicked items to be top-ranked. User preferences are learned using a Mallows ranking model. Bayesian inference leads to interpretable uncertainties of each individual recommendation, and we also propose a method to make personalized recommendations based on such uncertainties. With a simulation study and a real life data example, we demonstrate that compared to state-of-the-art matrix factorization, BMCD makes personalized recommendations with similar accuracy, while achieving much higher level of diversity, and producing interpretable and actionable uncertainty estimation.

stat.ME↗