Search

Methods for analyzing representations in neural systems are increasingly popular tools in neuroscience and mechanistic interpretability. Measures comparing neural activations across conditions, architectures, and species give scalable ways to understand information transformation within different neural networks. However, recent findings show that some metrics respond to spurious signals, leading to misleading results. Establishing benchmark test cases is thus essential for identifying the most reliable metric and potential improvements. We propose that compositional learning in recurrent neural networks (RNNs) can provide a test case for dynamical representation alignment metrics. Implementing this case allows us to evaluate if metrics can identify representations that develop throughout learning and determine if representations identified by metrics reflect the network's actual computations. Building both attractor and RNN based test cases, we show that the recently proposed Dynamical Similarity Analysis (DSA) is more noise robust and reliably identifies behaviorally relevant representations compared to prior metrics (Procrustes, CKA). We also demonstrate how such test cases can extend beyond metric evaluation to study new architectures. Specifically, testing DSA in modern (Mamba) state space models suggests that these models, unlike RNNs, may not require changes in recurrent dynamics due to their expressive hidden states. Overall, we develop test cases that showcase how DSA's enhanced ability to detect dynamical motifs makes it highly effective for identifying ongoing computations in RNNs and revealing how networks learn tasks.

Most of the intrusion detection methods in computer networks are based on traffic flow characteristics. However, this approach may not fully exploit the potential of deep learning algorithms to directly extract features and patterns from raw packets. Moreover, it impedes real-time monitoring due to the necessity of waiting for the processing pipeline to complete and introduces dependencies on additional software components. In this paper, we investigate deep learning methodologies capable of detecting attacks in real-time directly from raw packet data within network traffic. We propose a novel approach where packets are stacked into windows and separately recognised, with a 2D image representation suitable for processing with computer vision models. Our investigation utilizes the CIC IDS-2017 dataset, which includes both benign traffic and prevalent real-world attacks, providing a comprehensive foundation for our research., Comment: Submitted to Computer Science Journal

We consider preference disaggregation in the context of multiple criteria sorting. The value function parameters and thresholds separating the classes are inferred from the Decision Maker's (DM's) assignment examples. Given the multiplicity of sorting models compatible with indirect preferences, selecting a single, representative one can be conducted differently. We review several procedures for this purpose, aiming to identify the most discriminant, average, central, benevolent, aggressive, parsimonious, or robust models. Also, we present three novel procedures that implement the robust assignment rule in practice. They exploit stochastic acceptabilities and maximize the support given to the resulting assignments by all feasible sorting models. The performance of sixteen procedures is verified on problem instances with different complexities. The results of an experimental study indicate the most efficient procedure in terms of classification accuracy, reproducing the DM's model, and delivering the most robust assignments. These include approaches identifying differently interpreted centers of the feasible polyhedron and robust methods introduced in this paper. Moreover, we discuss how the performance of all procedures is affected by different numbers of classes, criteria, characteristic points, and reference assignments. Finally, we illustrate the use of all approaches in a study concerning the assessment of the green performance of European cities.

We announce and examine the conjecture that each infinite connected normal Hausdorff space has a quotient homeomorphic to the unit interval, shown to be true with the additional assumption of compactness or local connectedness. Some connected subsets of the plane are considered to show the difficulties involved in developing a general argument.

It is known that increasing powers of a continuous random variable converge in distribution to Benford's law as the exponent approaches infinity. The rate of convergence has been estimated using Fourier analysis, but we present an elementary method, which is easier to apply and provides a better estimation in the widely studied case of a uniformly distributed random variable., Comment: 6 pages

The main result of this paper states that for independent random variables $X, Y$ taking values in a compact metrisable abelian group, $X + Y$ has the same distribution as $X$, if and only if there exists a compact subgroup $A$ such that $P(Y\in A)=1$ and $X + a$ has the same distribution as $X$ for all $a\in A$. As a conclusion from the above it is shown that for independent random variables $X, Y$ such that $X+Y$ has the same distribution as $X$, $X+Y$ and $Y$ are also independent. It becomes also apparent that the distribution of $X$ is the Haar measure (uniform distribution) if for each open set $U$, $P(Y\in U) > 0$.

It is known that if X is uniformly distributed modulo 1 and Y is an arbitrary random variable independent of X then Y+X is also uniformly distributed modulo 1. We prove a converse for any continuous random variable Y (or a reasonable approximation to a continuous random variable) so that if X and Y+X are equally distributed modulo 1 and Y is independent of X then X is uniformly distributed modulo 1 (or approximates the uniform distribution equally reasonably). This translates into a characterization of Benford's law through a generalization of scale-invariance: from multiplication by a constant to multiplication by an independent random variable. We also show a base-invariance characterization: if a positive continuous random variable has the same significand distribution for two bases then it is Benford for both bases. The set of bases for which a random variable is Benford is characterized through characteristic functions., Comment: 25 pages

The main result of this paper states, that if a function $f:\R^2\to [0, +\infty)$ has a closed graph and the set of discontinuity points is a network (as defined by Kuratowski in Topology II, 61.IV), then the graph of $f$ is disconnected. It is also proven that this result can be easily generalised to a function $f:\R^2\to Y$ where $Y$ is a $\sigma$-locally compact metrisable space.

The main result of this paper states that for a function $f:\R^2\to Y$ with a closed, connected and locally connected graph, where $Y$ is a locally compact, second-countable metrisable space, the graph over discontinuity points remains locally connected.

We present a connected metric space that does not contain any nontrivial separable connected subspace. Our space is a dense connected graph of a function from the real line satisfying Cauchy's equation.