A simple example that I have been requested illustrates the statement in E-print nlin.CD/0201060 that solutions of a smooth first order dynamic equation can be made Lyapunov stable at will by the choice of an appropriate time-dependent Riemannian metric.

Department of Electrical Engineering, Pennsylvania State University, University Park, PA 16802

Online active learning is a paradigm in machine learning that aims to select the most informative data points to label from a data stream. The problem of minimizing the cost associated with collecting labeled observations has gained a lot of attention in recent years, particularly in real-world applications where data is only available in an unlabeled form. Annotating each observation can be time-consuming and costly, making it difficult to obtain large amounts of labeled data. To overcome this issue, many active learning strategies have been proposed in the last decades, aiming to select the most informative observations for labeling in order to improve the performance of machine learning models. These approaches can be broadly divided into two categories: static pool-based and stream-based active learning. Pool-based active learning involves selecting a subset of observations from a closed pool of unlabeled data, and it has been the focus of many surveys and literature reviews. However, the growing availability of data streams has led to an increase in the number of approaches that focus on online active learning, which involves continuously selecting and labeling observations as they arrive in a stream. This work aims to provide an overview of the most recently proposed approaches for selecting the most informative observations from data streams in real time. We review the various techniques that have been proposed and discuss their strengths and limitations, as well as the challenges and opportunities that exist in this area of research.

Identifying where quantum models may offer practical benefits in near term quantum machine learning (QML) requires moving beyond isolated algorithmic proposals toward systematic and empirical exploration across models, datasets, and hardware constraints. We introduce MerLin, an open-source framework designed as a discovery engine for photonic and hybrid quantum machine learning. MerLin integrates optimized strong simulation of linear optical circuits into standard PyTorch and scikit learn workflows, enabling end-to-end differentiable training of quantum layers. MerLin is designed around systematic benchmarking and reproducibility. As an initial contribution, we reproduce eighteen state-of-the-art photonic and hybrid QML works spanning kernel methods, reservoir computing, convolutional and recurrent architectures, generative models, and modern training paradigms. These reproductions are released as reusable, modular experiments that can be directly extended and adapted, establishing a shared experimental baseline consistent with empirical benchmarking methodologies widely adopted in modern artificial intelligence. By embedding photonic quantum models within established machine learning ecosystems, MerLin allows practitioners to leverage existing tooling for ablation studies, cross-modality comparisons, and hybrid classical-quantum workflows. The framework already implements hardware-aware features, allowing tests on available quantum hardware while enabling exploration beyond its current capabilities, positioning MerLin as a forward-looking co-design tool linking algorithms, benchmarks, and hardware.

This article provides the first survey of computational models of emotion in reinforcement learning (RL) agents. The survey focuses on agent/robot emotions, and mostly ignores human user emotions. Emotions are recognized as functional in decision-making by influencing motivation and action selection. Therefore, computational emotion models are usually grounded in the agent's decision making architecture, of which RL is an important subclass. Studying emotions in RL-based agents is useful for three research fields. For machine learning (ML) researchers, emotion models may improve learning efficiency. For the interactive ML and human-robot interaction (HRI) community, emotions can communicate state and enhance user investment. Lastly, it allows affective modelling (AM) researchers to investigate their emotion theories in a successful AI agent class. This survey provides background on emotion theory and RL. It systematically addresses 1) from what underlying dimensions (e.g., homeostasis, appraisal) emotions can be derived and how these can be modelled in RL-agents, 2) what types of emotions have been derived from these dimensions, and 3) how these emotions may either influence the learning efficiency of the agent or be useful as social signals. We also systematically compare evaluation criteria, and draw connections to important RL sub-domains like (intrinsic) motivation and model-based RL. In short, this survey provides both a practical overview for engineers wanting to implement emotions in their RL agents, and identifies challenges and directions for future emotion-RL research.

We seek to enable classic processing of continuous ultra-sparse spatiotemporal data generated by event-based sensors with dense machine learning models. We propose a novel hybrid pipeline composed of asynchronous sensing and synchronous processing that combines several ideas: (1) an embedding based on PointNet models -- the ALERT module -- that can continuously integrate new and dismiss old events thanks to a leakage mechanism, (2) a flexible readout of the embedded data that allows to feed any downstream model with always up-to-date features at any sampling rate, (3) exploiting the input sparsity in a patch-based approach inspired by Vision Transformer to optimize the efficiency of the method. These embeddings are then processed by a transformer model trained for object and gesture recognition. Using this approach, we achieve performances at the state-of-the-art with a lower latency than competitors. We also demonstrate that our asynchronous model can operate at any desired sampling rate.

Automatic machine learning is an important problem in the forefront of machine learning. The strongest AutoML systems are based on neural networks, evolutionary algorithms, and Bayesian optimization. Recently AlphaD3M reached state-of-the-art results with an order of magnitude speedup using reinforcement learning with self-play. In this work we extend AlphaD3M by using a pipeline grammar and a pre-trained model which generalizes from many different datasets and similar tasks. Our results demonstrate improved performance compared with our earlier work and existing methods on AutoML benchmark datasets for classification and regression tasks. In the spirit of reproducible research we make our data, models, and code publicly available.

We present the results of processing the effects of the powerful Gamma Ray Burst GRB221009A captured by the charged particle detectors (electrostatic analyzers and solid-state detectors) onboard spacecraft at different points in the heliosphere on October 9, 2022. To follow the GRB221009A propagation through the heliosphere we used the electron and proton flux measurements from solar missions Solar Orbiter and STEREO-A; Earth magnetosphere and the solar wind missions THEMIS and Wind; meteorological satellites POES15, POES19, MetOp3; and MAVEN - a NASA mission orbiting Mars. GRB221009A had a structure of four bursts: less intense Pulse 1 - the triggering impulse - was detected by gamma-ray observatories at 131659 UT (near the Earth); the most intense Pulses 2 and 3 were detected on board all the spacecraft from the list, and Pulse 4 detected in more than 500 s after Pulse 1. Due to their different scientific objectives, the spacecraft, which data was used in this study, were separated by more than 1 AU (Solar Orbiter and MAVEN). This enabled tracking GRB221009A as it was propagating across the heliosphere. STEREO-A was the first to register Pulse 2 and 3 of the GRB, almost 100 seconds before their detection by spacecraft in the vicinity of Earth. MAVEN detected GRB221009A Pulses 2, 3, and 4 at the orbit of Mars about 237 seconds after their detection near Earth. By processing the time delays observed we show that the source location of the GRB221009A was at RA 288.5 degrees, Dec 18.5 degrees (J2000) with an error cone of 2 degrees

We argue that the fundamental Theory of Everything is a conventional field theory defined in the flat multidimensional bulk. Our Universe should be obtained as a 3-brane classical solution in this theory. The renormalizability of the fundamental theory implies that it involves higher derivatives (HD). It should be supersymmetric (otherwise one cannot get rid of the huge induced cosmological term) and probably conformal (otherwise one can hardly cope with the problem of ghosts) . We present arguments that in conformal HD theories the ghosts (which are inherent for HD theories) might be not so malignant. In particular, we present a nontrivial QM HD model where ghosts are absent and the spectrum has a well defined ground state. The requirement of superconformal invariance restricts the dimension of the bulk to be D < 7. We suggest that the TOE lives in six dimensions and enjoys the maximum N = (2,0) superconformal symmetry. Unfortunately, no renormalizable field theory with this symmetry is presently known. We construct and discuss an N = (1,0) 6D supersymmetric gauge theory with four derivatives in the action. This theory involves a dimensionless coupling constant and is renormalizable. At the tree level, the theory enjoys conformal symmetry, but the latter is broken by quantum anomaly. The sign of the beta function corresponds to the Landau zero situation.

The folding pathway and rate coefficients of the folding of a knotted protein are calculated for a potential energy function with minimal energetic frustration. A kinetic transition network is constructed using the discrete path sampling approach, and the resulting potential energy surface is visualized by constructing disconnectivity graphs. Owing to topological constraints, the low-lying portion of the landscape consists of three distinct regions, corresponding to the native knotted state and to configurations where either the N- or C-terminus is not yet folded into the knot. The fastest folding pathways from denatured states exhibit early formation of the N-terminus portion of the knot and a rate-determining step where the C-terminus is incorporated. The low-lying minima with the N-terminus knotted and the C-terminus free therefore constitute an off-pathway intermediate for this model. The insertion of both the N- and C-termini into the knot occur late in the folding process, creating large energy barriers that are the rate limiting steps in the folding process. When compared to other protein folding proteins of a similar length, this system folds over six orders of magnitude more slowly.