arXiv subjects

Miguel Romero

Publications and source records attributed to Miguel Romero.

At least 19 recordsLinked to original sources

arXiv2025

Beyond instruction-conditioning, MoTE: Mixture of Task Experts for Multi-task Embedding Models

Dense embeddings are fundamental to modern machine learning systems, powering Retrieval-Augmented Generation (RAG), information retrieval, and representation learning. While instruction-conditioning has become the dominant approach for embedding specialization, its direct application to low-capacity models imposes fundamental representational constraints that limit the performance gains derived from specialization. In this paper, we analyze these limitations and introduce the Mixture of Task Experts (MoTE) transformer block, which leverages task-specialized parameters trained with Task-Aware Contrastive Learning (\tacl) to enhance the model ability to generate specialized embeddings. Empirical results show that MoTE achieves $64\%$ higher performance gains in retrieval datasets ($+3.27 \rightarrow +5.21$) and $43\%$ higher performance gains across all datasets ($+1.81 \rightarrow +2.60$). Critically, these gains are achieved without altering instructions, training data, inference time, or number of active parameters.

cs.LG↗

arXiv2025

Minimizing Conjunctive Regular Path Queries

We study the minimization problem for Conjunctive Regular Path Queries (CRPQs) and unions of CRPQs (UCRPQs). This is the problem of checking, given a query and a number $k$, whether the query is equivalent to one of size at most $k$. For CRPQs we consider the size to be the number of atoms, and for UCRPQs the maximum number of atoms in a CRPQ therein, motivated by the fact that the number of atoms has a leading influence on the cost of query evaluation. We show that the minimization problem is decidable, both for CRPQs and UCRPQs. We provide a 2ExpSpace upper-bound for CRPQ minimization, based on a brute-force enumeration algorithm, and an ExpSpace lower-bound. For UCRPQs, we show that the problem is ExpSpace-complete, having thus the same complexity as the classical containment problem. The upper bound is obtained by defining and computing a notion of maximal under-approximation. Moreover, we show that for UCRPQs using the so-called "simple regular expressions" consisting of concatenations of expressions of the form $a^+$ or $a_1 + \dotsb + a_k$, the minimization problem becomes $Π^p_2$-complete, again matching the complexity of containment.

cs.DB↗

arXiv2024

The Distributional Uncertainty of the SHAP score in Explainable Machine Learning

Attribution scores reflect how important the feature values in an input entity are for the output of a machine learning model. One of the most popular attribution scores is the SHAP score, which is an instantiation of the general Shapley value used in coalition game theory. The definition of this score relies on a probability distribution on the entity population. Since the exact distribution is generally unknown, it needs to be assigned subjectively or be estimated from data, which may lead to misleading feature scores. In this paper, we propose a principled framework for reasoning on SHAP scores under unknown entity population distributions. In our framework, we consider an uncertainty region that contains the potential distributions, and the SHAP score of a feature becomes a function defined over this region. We study the basic problems of finding maxima and minima of this function, which allows us to determine tight ranges for the SHAP scores of all features. In particular, we pinpoint the complexity of these problems, and other related ones, showing them to be NP-complete. Finally, we present experiments on a real-world dataset, showing that our framework may contribute to a more robust feature scoring.

cs.AI↗

arXiv2024

A Neuro-Symbolic Framework for Answering Graph Pattern Queries in Knowledge Graphs

The challenge of answering graph queries over incomplete knowledge graphs is gaining significant attention in the machine learning community. Neuro-symbolic models have emerged as a promising approach, combining good performance with high interpretability. These models utilize trained architectures to execute atomic queries and integrate modules that mimic symbolic query operators. However, most neuro-symbolic query processors are constrained to tree-like graph pattern queries. These queries admit a bottom-up execution with constant values or anchors at the leaves and the target variable at the root. While expressive, tree-like queries fail to capture critical properties in knowledge graphs, such as the existence of multiple edges between entities or the presence of triangles. We introduce a framework for answering arbitrary graph pattern queries over incomplete knowledge graphs, encompassing both cyclic queries and tree-like queries with existentially quantified leaves. These classes of queries are vital for practical applications but are beyond the scope of most current neuro-symbolic models. Our approach employs an approximation scheme that facilitates acyclic traversals for cyclic patterns, thereby embedding additional symbolic bias into the query execution process. Our experimental evaluation demonstrates that our framework performs competitively on three datasets, effectively handling cyclic queries through our approximation strategy. Additionally, it maintains the performance of existing neuro-symbolic models on anchored tree-like queries and extends their capabilities to queries with existentially quantified variables.

cs.LG↗

arXiv2023

Pliability and Approximating Max-CSPs

We identify a sufficient condition, treewidth-pliability, that gives a polynomial-time algorithm for an arbitrarily good approximation of the optimal value in a large class of Max-2-CSPs parameterised by the class of allowed constraint graphs (with arbitrary constraints on an unbounded alphabet). Our result applies more generally to the maximum homomorphism problem between two rational-valued structures. The condition unifies the two main approaches for designing a polynomial-time approximation scheme. One is Baker's layering technique, which applies to sparse graphs such as planar or excluded-minor graphs. The other is based on Szemerédi's regularity lemma and applies to dense graphs. We extend the applicability of both techniques to new classes of Max-CSPs. On the other hand, we prove that the condition cannot be used to find solutions (as opposed to approximating the optimal value) in general. Treewidth-pliability turns out to be a robust notion that can be defined in several equivalent ways, including characterisations via size, treedepth, or the Hadwiger number. We show connections to the notions of fractional-treewidth-fragility from structural graph theory, hyperfiniteness from the area of property testing, and regularity partitions from the theory of dense graph limits. These may be of independent interest. In particular we show that a monotone class of graphs is hyperfinite if and only if it is fractionally-treewidth-fragile and has bounded degree.

cs.DM↗

arXiv2023

Conjunctive Regular Path Queries under Injective Semantics

We introduce injective semantics for Conjunctive Regular Path Queries (CRPQs), and study their fundamental properties. We identify two such semantics: atom-injective and query-injective semantics, both defined in terms of injective homomorphisms. These semantics are natural generalizations of the well-studied class of RPQs under simple-path semantics to the class of CRPQs. We study their evaluation and containment problems, providing useful characterizations for them, and we pinpoint the complexities of these problems. Perhaps surprisingly, we show that containment for CRPQs becomes undecidable for atom-injective semantics, and PSPACE-complete for query-injective semantics, in contrast to the known EXPSPACE-completeness result for the standard semantics. The techniques used differ significantly from the ones known for the standard semantics, and new tools tailored to injective semantics are needed. We complete the picture of complexity by investigating, for each semantics, the containment problem for the main subclasses of CRPQs, namely Conjunctive Queries and CRPQs with finite languages.

cs.DB↗

arXiv2022

A Rewriting Logic Approach to Stochastic and Spatial Constraint System Specification and Verification

This paper addresses the issue of specifying, simulating, and verifying reactive systems in rewriting logic. It presents an executable semantics for probabilistic, timed, and spatial concurrent constraint programming -- here called stochastic and spatial concurrent constraint systems (SSCC) -- in the rewriting logic semantic framework. The approach is based on an enhanced and generalized model of concurrent constraint programming (CCP) where computational hierarchical spaces can be assigned to belong to agents. The executable semantics faithfully represents and operationally captures the highly concurrent nature, uncertain behavior, and spatial and epistemic characteristics of reactive systems with flow of information. In SSCC, timing attributes -- represented by stochastic duration -- can be associated to processes, and exclusive and independent probabilistic choice is also supported. SMT solving technology, available from the Maude system, is used to realize the underlying constraint system of SSCC with quantifier-free formulas over integers and reals. This results in a fully executable real-time symbolic specification that can be used for quantitative analysis in the form of statistical model checking. The main features and capabilities of SSCC are illustrated with examples throughout the paper. This contribution is part of a larger research effort aimed at making available formal analysis techniques and tools, mathematically founded on the CCP approach, to the research community.

cs.LO↗

arXiv2022

Hierarchy exploitation to detect missing annotations on hierarchical multi-label classification

The availability of genomic data has grown exponentially in the last decade, mainly due to the development of new sequencing technologies. Based on the interactions between genes (and gene products) extracted from the increasing genomic data, numerous studies have focused on the identification of associations between genes and functions. While these studies have shown great promise, the problem of annotating genes with functions remains an open challenge. In this work, we present a method to detect missing annotations in hierarchical multi-label classification datasets. We propose a method that exploits the class hierarchy by computing aggregated probabilities to the paths of classes from the leaves to the root for each instance. The proposed method is presented in the context of predicting missing gene function annotations, where these aggregated probabilities are further used to select a set of annotations to be verified through in vivo experiments. The experiments on Oriza sativa Japonica, a variety of rice, showcase that incorporating the hierarchy of classes into the method often improves the predictive performance and our proposed method yields superior results when compared to competitor methods from the literature.

cs.LG↗

arXiv2022

On Computing Probabilistic Explanations for Decision Trees

Formal XAI (explainable AI) is a growing area that focuses on computing explanations with mathematical guarantees for the decisions made by ML models. Inside formal XAI, one of the most studied cases is that of explaining the choices taken by decision trees, as they are traditionally deemed as one of the most interpretable classes of models. Recent work has focused on studying the computation of "sufficient reasons", a kind of explanation in which given a decision tree $T$ and an instance $x$, one explains the decision $T(x)$ by providing a subset $y$ of the features of $x$ such that for any other instance $z$ compatible with $y$, it holds that $T(z) = T(x)$, intuitively meaning that the features in $y$ are already enough to fully justify the classification of $x$ by $T$. It has been argued, however, that sufficient reasons constitute a restrictive notion of explanation, and thus the community has started to study their probabilistic counterpart, in which one requires that the probability of $T(z) = T(x)$ must be at least some value $δ\in (0, 1]$, where $z$ is a random instance that is compatible with $y$. Our paper settles the computational complexity of $δ$-sufficient-reasons over decision trees, showing that both (1) finding $δ$-sufficient-reasons that are minimal in size, and (2) finding $δ$-sufficient-reasons that are minimal inclusion-wise, do not admit polynomial-time algorithms (unless P=NP). This is in stark contrast with the deterministic case ($δ= 1$) where inclusion-wise minimal sufficient-reasons are easy to compute. By doing this, we answer two open problems originally raised by Izza et al. On the positive side, we identify structural restrictions of decision trees that make the problem tractable, and show how SAT solvers might be able to tackle these problems in practical settings.

cs.LG↗

arXiv2022

Modeling GPU Dynamic Parallelism for Self Similar Density Workloads

Dynamic Parallelism (DP) is a runtime feature of the GPU programming model that allows GPU threads to execute additional GPU kernels, recursively. Apart from making the programming of parallel hierarchical patterns easier, DP can also speedup problems that exhibit a heterogeneous data layout by focusing, through a subdivision process, the finite GPU resources on the sub-regions that exhibit more parallelism. However, doing an optimal subdivision process is not trivial, as there are different parameters that play an important role in the final performance of DP. Moreover, the current programming abstraction for DP also introduces an overhead that can penalize the final performance. In this work we present a subdivision cost model for problems that exhibit self similar density (SSD) workloads (such as fractals), in order understand what parameters provide the fastest subdivision approach. Also, we introduce a new subdivision implementation, named \textit{Adaptive Serial Kernels} (ASK), as a smaller overhead alternative to CUDA's Dynamic Parallelism. Using the cost model on the Mandelbrot Set as a case study shows that the optimal scheme is to start with an initial subdivision between $g=[2,16]$, then keep subdividing in regions of $r=2,4$, and stop when regions reach a size of $B \sim 32$. The experimental results agree with the theoretical parameters, confirming the usability of the cost model. In terms of performance, the proposed ASK approach runs up to $\sim 60\%$ faster than Dynamic Parallelism in the Mandelbrot set, and up to $12\times$ faster than a basic exhaustive implementation, whereas DP is up to $7.5\times$.

cs.DC↗

arXiv2022

Feature extraction using Spectral Clustering for Gene Function Prediction using Hierarchical Multi-label Classification

Gene annotation addresses the problem of predicting unknown associations between gene and functions (e.g., biological processes) of a specific organism. Despite recent advances, the cost and time demanded by annotation procedures that rely largely on in vivo biological experiments remain prohibitively high. This paper presents a novel in silico approach for to the annotation problem that combines cluster analysis and hierarchical multi-label classification (HMC). The approach uses spectral clustering to extract new features from the gene co-expression network (GCN) and enrich the prediction task. HMC is used to build multiple estimators that consider the hierarchical structure of gene functions. The proposed approach is applied to a case study on Zea mays, one of the most dominant and productive crops in the world. The results illustrate how in silico approaches are key to reduce the time and costs of gene annotation. More specifically, they highlight the importance of: (i) building new features that represent the structure of gene relationships in GCNs to annotate genes; and (ii) taking into account the structure of biological processes to obtain consistent predictions.

cs.LG↗

arXiv2022

A Top-down Supervised Learning Approach to Hierarchical Multi-label Classification in Networks

Node classification is the task of inferring or predicting missing node attributes from information available for other nodes in a network. This paper presents a general prediction model to hierarchical multi-label classification (HMC), where the attributes to be inferred can be specified as a strict poset. It is based on a top-down classification approach that addresses hierarchical multi-label classification with supervised learning by building a local classifier per class. The proposed model is showcased with a case study on the prediction of gene functions for Oryza sativa Japonica, a variety of rice. It is compared to the Hierarchical Binomial-Neighborhood, a probabilistic model, by evaluating both approaches in terms of prediction performance and computational cost. The results in this work support the working hypothesis that the proposed model can achieve good levels of prediction efficiency, while scaling up in relation to the state of the art.

cs.LG↗

arXiv2021

The complexity of general-valued CSPs seen from the other side

The constraint satisfaction problem (CSP) is concerned with homomorphisms between two structures. For CSPs with restricted left-hand side structures, the results of Dalmau, Kolaitis, and Vardi [CP'02], Grohe [FOCS'03/JACM'07], and Atserias, Bulatov, and Dalmau [ICALP'07] establish the precise borderline of polynomial-time solvability (subject to complexity-theoretic assumptions) and of solvability by bounded-consistency algorithms (unconditionally) as bounded treewidth modulo homomorphic equivalence. The general-valued constraint satisfaction problem (VCSP) is a generalisation of the CSP concerned with homomorphisms between two valued structures. For VCSPs with restricted left-hand side valued structures, we establish the precise borderline of polynomial-time solvability (subject to complexity-theoretic assumptions) and of solvability by the $k$-th level of the Sherali-Adams LP hierarchy (unconditionally). We also obtain results on related problems concerned with finding a solution and recognising the tractable cases; the latter has an application in database theory.

cs.CC↗

arXiv2020

Targeted transfer learning to improve performance in small medical physics datasets

The growing use of Machine Learning has produced significant advances in many fields. For image-based tasks, however, the use of deep learning remains challenging in small datasets. In this article, we review, evaluate and compare the current state-of-the-art techniques in training neural networks to elucidate which techniques work best for small datasets. We further propose a path forward for the improvement of model accuracy in medical imaging applications. We observed best results from one cycle training, discriminative learning rates with gradual freezing and parameter modification after transfer learning. We also established that when datasets are small, transfer learning plays an important role beyond parameter initialization by reusing previously learned features. Surprisingly we observed that there is little advantage in using pre-trained networks in images from another part of the body compared to Imagenet. On the contrary, if images from the same part of the body are available then transfer learning can produce a significant improvement in performance with as little as 50 images in the training data.

cs.LG↗

arXiv2020

Spectral Evolution with Approximated Eigenvalue Trajectories for Link Prediction

The spectral evolution model aims to characterize the growth of large networks (i.e., how they evolve as new edges are established) in terms of the eigenvalue decomposition of the adjacency matrices. It assumes that, while eigenvectors remain constant, eigenvalues evolve in a predictable manner over time. This paper extends the original formulation of the model twofold. First, it presents a method to compute an approximation of the spectral evolution of eigenvalues based on the Rayleigh quotient. Second, it proposes an algorithm to estimate the evolution of eigenvalues by extrapolating only a fraction of their approximated values. The proposed model is used to characterize mention networks of users who posted tweets that include the most popular political hashtags in Colombia from August 2017 to August 2018 (the period which concludes the disarmament of the Revolutionary Armed Forces of Colombia). To evaluate the extent to which the spectral evolution model resembles these networks, link prediction methods based on learning algorithms (i.e., extrapolation and regression) and graph kernels are implemented. Experimental results show that the learning algorithms deployed on the approximated trajectories outperform the usual kernel and extrapolation methods at predicting the formation of new edges.

cs.LG↗

arXiv2020

On monotonic determinacy and rewritability for recursive queries and views

A query Q is monotonically determined over a set of views if Q can be expressed as a monotonic function of the view image. In the case of relational algebra views and queries, monotonic determinacy coincides with rewritability as a union of conjunctive queries, and it is decidable in important special cases, such as for CQ views and queries. We investigate the situation for views and queries in the recursive query language Datalog. We give both positive and negative results about the ability to decide monotonic determinacy, and also about the co-incidence of monotonic determinacy with Datalog rewritability.

cs.LO↗

arXiv2019

A tractable class of binary VCSPs via M-convex intersection

A binary VCSP is a general framework for the minimization problem of a function represented as the sum of unary and binary cost functions. An important line of VCSP research is to investigate what functions can be solved in polynomial time. Cooper and Živný classified the tractability of binary VCSP instances according to the concept of "triangle," and showed that the only interesting tractable case is the one induced by the joint winner property (JWP). Recently, Iwamasa, Murota, and Živný made a link between VCSP and discrete convex analysis, showing that a function satisfying the JWP can be transformed into a function represented as the sum of two quadratic M-convex functions, which can be minimized in polynomial time via an M-convex intersection algorithm if the value oracle of each M-convex function is given. In this paper, we give an algorithmic answer to a natural question: What binary finite-valued CSP instances can be represented as the sum of two quadratic M-convex functions and can be solved in polynomial time via an M-convex intersection algorithm? We solve this problem by devising a polynomial-time algorithm for obtaining a concrete form of the representation in the representable case. Our result presents a larger tractable class of binary finite-valued CSPs, which properly contains the JWP class.

cs.DM↗

arXiv2019

Boundedness of Conjunctive Regular Path Queries

We study the boundedness problem for unions of conjunctive regular path queries with inverses (UC2RPQs). This is the problem of, given a UC2RPQ, checking whether it is equivalent to a union of conjunctive queries (UCQ). We show the problem to be ExpSpace-complete, thus coinciding with the complexity of containment for UC2RPQs. As a corollary, when a UC2RPQ is bounded, it is equivalent to a UCQ of at most triple-exponential size, and in fact we show that this bound is optimal. We also study better behaved classes of UC2RPQs, namely acyclic UC2RPQs of bounded thickness, and strongly connected UCRPQs, whose boundedness problem are, respectively, PSpace-complete and $Π^p_2$-complete. Most upper bounds exploit results on limitedness for distance automata, in particular extending the model with alternation and two-wayness, which may be of independent interest.

cs.DB↗