This study investigates the application of various artificial neural network (ANN) architectures for music genre classification, focusing on the evolution of models and their performance on the GTZAN dataset. Through experimentation with fully connected networks, convolutional neural networks (CNNs), recurrent neural networks (RNNs), and Transformer-based architectures, we analyze their strengths and limitations in genre classification. The results demonstrate that hyperparameter tuning, dataset diversity, and model selection significantly impact classification performance. Although CNNs, RNNs, and gated recurrent units (GRUs) achieve more precision than 90% in evaluation tests, the addition of Transformer layers does not consistently improve classification accuracy and may exacerbate challenges in certain genres. These results underscore the importance of carefully considering model architecture and dataset characteristics in music genre classification. Future research should focus on developing extensive, multi-rater verified datasets to enhance classification performance and model robustness in music genre classification tasks.

Neural network pruning is a popular approach to reducing model storage size and inference time by removing redundant parameters in the neural network. However, the uncertainty of predictions from pruned models is unexplored. In this paper we study neural network pruning in the context of conformal predictors (CP). The conformal prediction framework built on top of machine learning algorithms supplements their predictions with reliable uncertainty measure in the form of prediction sets, under the independent and identically distributed assumption on the data. Convolutional neural networks (CNNs) have complicated architectures and are widely used in various applications nowadays. Therefore, we focus on pruning CNNs and, in particular, filter-level pruning. We first propose a brute force method that estimates the contribution of a filter to the CP’s predictive efficiency and removes those with the least contribution. Given the computation inefficiency of the brute force method, we also propose the Taylor expansion to approximate the filter’s contribution. Furthermore, we improve the global pruning method by protecting the most important filters within each layer from being pruned. In addition, we explore the ConfTr loss function which is optimized to yield maximal CP efficiency in the context of neural network pruning. We have conducted extensive experimental studies and compared the results regarding the trade-offs between predictive efficiency, computational efficiency, and network sparsity. These results are instructive for deploying pruned neural networks with applications using conformal prediction where reliable predictions and reduced computational cost are relevant, such as in safety-critical applications.

Portfolio optimization is a strategic approach aiming at achieving an optimal balance between risk and returns through the judicious allocation of limited capital across various assets. In recent years, there has been a growing interest in leveraging Deep Reinforcement Learning (DRL) to tackle the complexities of portfolio optimization. Despite its potential, a notable limitation of DRL algorithms is their inherent difficulty in integrating conflicted objectives with the reward functions throughout the learning process. Typically, DRL's reward function prioritizes the maximization of returns or other performance indicators, often overlooking the integration of risk aspects. Furthermore, the standard DRL framework struggles to incorporate practical constraints, such as cardinality and bounding, into the decision process. Without these constraints, the investment strategies developed might be unrealistic and unmanageable. To this end, in this paper, we propose an adaptive and safe DRL framework, which can dynamically optimize the portfolio weights while strictly respecting practical constraints. In our method, any infeasible action (i.e., one that violates the constraints) decided by the RL agent will be mapped to a feasible region using a safety layer. The extended Markowitz Mean-Variance (M-V) model is explicitly encoded in the safety layer to ensure the feasibility of the actions from the alternative views. In addition, we utilize Projection-based Interior-point Policy Optimization (IPO) to resolve multiple objectives and constraints in the examined problem. Extensive results on real-world datasets show that our method is effective in strictly respecting constraints under dynamic market environments, in contrast to prevailing data- driven trading strategies and conventional model-based static solutions.

We propose a logic of east and west (LEW ) for points in 1D Euclidean space. It formalises primitive direction relations: east (E), west (W) and indeterminate east/west (Iew). It has a parameter τ ∈ N>1, which is referred to as the level of indeterminacy in directions. For every τ ∈ N>1, we provide a sound and complete axiomatisation of LEW , and prove that its satisfiability problem is NP-complete. In addition, we show that the finite axiomatisability of LEW depends on τ : if τ = 2 or τ = 3, then there exists a finite sound and complete axiomatisation; if τ > 3, then the logic is not finitely axiomatisable. LEW can be easily extended to higher-dimensional Euclidean spaces. Extending LEW to 2D Euclidean space makes it suitable for reasoning about not perfectly aligned representations of the same spatial objects in different datasets, for example, in crowd-sourced digital maps.

Abstract Robustness is a property of system analyses, namely monotonic maps from the complete lattice of subsets of a (system’s state) space to the two-point lattice. The definition of robustness requires the space to be a metric space. Robust analyses cannot discriminate between a subset of the metric space and its closure; therefore, one can restrict to the complete lattice of closed subsets. When the metric space is compact, the complete lattice of closed subsets ordered by reverse inclusion is $\omega$ -continuous, and robust analyses are exactly the Scott-continuous maps. Thus, one can also ask whether a robust analysis is computable (with respect to a countable base). The main result of this paper establishes a relation between robustness and Scott continuity when the metric space is not compact. The key idea is to replace the metric space with a compact Hausdorff space, and relate robustness and Scott continuity by an adjunction between the complete lattice of closed subsets of the metric space and the $\omega$ -continuous lattice of closed subsets of the compact Hausdorff space. We demonstrate the applicability of this result with several examples involving Banach spaces.

Abstract A domain-theoretic framework is presented for validated robustness analysis of neural networks. First, global robustness of a general class of networks is analyzed. Then, using the fact that Edalat’s domain-theoretic L -derivative coincides with Clarke’s generalized gradient, the framework is extended for attack-agnostic local robustness analysis. The proposed framework is ideal for designing algorithms which are correct by construction. This claim is exemplified by developing a validated algorithm for estimation of Lipschitz constant of feedforward regressors. The completeness of the algorithm is proved over differentiable networks and also over general position ${\mathrm{ReLU}}$ networks. Computability results are obtained within the framework of effectively given domains. Using the proposed domain model, differentiable and non-differentiable networks can be analyzed uniformly. The validated algorithm is implemented using arbitrary-precision interval arithmetic, and the results of some experiments are presented. The software implementation is truly validated, as it handles floating-point errors as well.

We construct a continuous domain, as a model of interval analysis, for temporal discretization of differential equations. By using this domain, and the domain of Lipschitz maps, we formulate a generalization of the Euler operator, which exhibits second-order convergence. We prove computability of the operator within the framework of effectively given domains. The operator only requires the vector field of the differential equation to be Lipschitz continuous, in contrast to the related operators in the literature which require the vector field to be at least continuously differentiable. Within the same framework, we also analyze temporal discretization and computability of another variant of the Euler operator formulated according to Runge-Kutta theory. We prove that, compared with this variant, the second-order operator that we formulate directly, not only imposes weaker assumptions on the vector field, but also exhibits superior convergence rate. We implement the first-order, second-order, and Runge-Kutta Euler operators using arbitrary-precision interval arithmetic, and report on some experiments. The experiments confirm our theoretical results. In particular, we observe the superior convergence rate of our second-order operator compared with the Runge-Kutta Euler and the common (first-order) Euler operators.

Transportation Electrification and Micro-grid development have raised the requirements for the DC-DC converter that performs the bus interface, the renewable energy and storage integration. The Multiple Active Bridge converter allows for the independent power transfer between its ports while keeping the electrical isolation and offering soft-switching behavior when operated with a phase-shift modulation; however, its performance is affected by the high level of coupling between the ports. In fact, the input voltage, or the phase shift of one port affects the power flow in the whole converter. This paper proposes an Artificial Neural Network (ANN) based decoupling that allows for an improved individual power regulation between the ports, considering the non-linear behavior of the system and the effects of the voltage variations.

We propose a method for generating music from a given image through three stages of translation, from image to caption, caption to lyrics, and lyrics to instrumental music, which forms the content to be combined with a given style. We train our proposed model, which we call BGT (BLIP-GPT2-TeleMelody), on two open-source datasets, one containing over 200,000 labeled images, and another containing more than 175,000 MIDI music files. In contrast with pixel level translation, the BGT model retains the semantics of the input image. We verify our claim through a user study in which participants were asked to match input images with generated music without access to the intermediate caption and lyrics. The results show that, while the matching rate among participants with low music expertise is essentially random, the rate among those with composition experience is significantly high, which strongly indicates that some semantic content of the input image is retained in the generated music.

A domain-theoretic method for solving initial value problems (IVPs) is presented, together with proofs of soundness, completeness, and some results on the algebraic complexity of the method. While the common fixed-precision interval arithmetic methods are restricted by the precision of the underlying machine architecture, domain-theoretic methods may be complete, i.e., the result may be obtained to any degree of accuracy. Furthermore, unlike methods based on interval arithmetic which require access to the syntactic representation of the vector field, domain-theoretic methods only deal with the semantics of the field, in the sense that the field is assumed to be given via finitely-representable approximations, to within any required accuracy. In contrast to the domain-theoretic first-order Euler method, the second-order method uses the local Lipschitz properties of the field. This is achieved by using a domain for Lipschitz functions, whose elements are consistent pairs that provide approximations of the field and its local Lipschitz properties. In the special case where the field is differentiable, the local Lipschitz properties are exactly the local differential properties of the field. In solving IVPs, Lipschitz continuity of the field is a common assumption, as a sufficient condition for uniqueness of the solution. While the validated methods for solving IVPs commonly impose further restrictions on the vector field, the second-order Euler method requires no further condition. In this sense, the method may be seen as the most general of its kind. To avoid complicated notations and lengthy arguments, the results of the paper are stated for the second-order Euler method. Nonetheless, the framework, and the results, may be extended to any higher-order Euler method, in a straightforward way.

Hybrid systems—more precisely, their mathematical models—can exhibit behaviors, like Zeno behaviors, that are absent in purely discrete or purely continuous systems. First, we observe that, in this context, the usual definition of reachability—namely, the reflexive and transitive closure of a transition relation—can be unsafe, i.e., it may compute a proper subset of the set of states reachable in finite time from a set of initial states. Therefore, we propose safe reachability, which always computes a superset of the set of reachable states. Second, in safety analysis of hybrid and continuous systems, it is important to ensure that a reachability analysis is also robust w.r.t. small perturbations to the set of initial states and to the system itself, since discrepancies between a system and its mathematical models are unavoidable. We show that, under certain conditions, the best Scott continuous approximation of an analysis A is also its best robust approximation. Finally, we exemplify the gap between the set of reachable states and the supersets computed by safe reachability and its best robust approximation.