Search

Recent research has established a connection between modern Hopfield networks (HNs) and transformer attention heads, with guarantees of exponential storage capacity. However, these models still face challenges scaling storage efficiently. Inspired by psychological theories of continuous neural resource allocation in working memory, we propose an approach that compresses large discrete Hopfield memories into smaller, continuous-time memories. Leveraging continuous attention, our new energy function modifies the update rule of HNs, replacing the traditional softmax-based probability mass function with a probability density, over the continuous memory. This formulation aligns with modern perspectives on human executive function, offering a principled link between attractor dynamics in working memory and resource-efficient memory allocation. Our framework maintains competitive performance with HNs while leveraging a compressed memory, reducing computational costs across synthetic and video datasets.

Current video-language models struggle with long-video understanding due to limited context lengths and reliance on sparse frame subsampling, often leading to information loss. This paper introduces $\infty$-Video, which can process arbitrarily long videos through a continuous-time long-term memory (LTM) consolidation mechanism. Our framework augments video Q-formers by allowing them to process unbounded video contexts efficiently and without requiring additional training. Through continuous attention, our approach dynamically allocates higher granularity to the most relevant video segments, forming "sticky" memories that evolve over time. Experiments with Video-LLaMA and VideoChat2 demonstrate improved performance in video question-answering tasks, showcasing the potential of continuous-time LTM mechanisms to enable scalable and training-free comprehension of long videos., Comment: 17 pages, 7 figures

Y$_2$Co$_3$ is a newly discovered antiferromagnetic (AFM) compound with distorted kagome layers. Previous investigations via bulk magnetization measurements suggested a complex noncollinear magnetic behavior, with magnetic moments primarily anti-aligned along the $b$ axis and some canting towards the $ac$ plane. In this study, we report the magnetic structure of Y$_2$Co$_3$ to be an A-type AFM structure with ferromagnetic (FM) interactions within the distorted kagome plane and an interplane antiferromagnetic interaction, as determined by single-crystal neutron diffraction. The magnetic moments align along the $b$ axis, with minimal canting towards the $c$ axis, at odds with the previous interpretation of bulk magnetization measurements. The magnetic moments on the two distinct Co sites are [0, -0.68(3), 0] $\mu_B$ and [0, 1.25(4), 0.07(1)] $\mu_B$. We attribute the previously reported "noncollinear" behavior to the considerable temperature dependence of itinerant AFM exchange interactions, induced by thermal contraction along the $b$ axis. Additionally, our examination of lattice constants through pressure studies reveals compensating effects on FM and AFM interactions, resulting in negligible pressure dependence of $T_\textrm{N}$.

Associative memory models, such as Hopfield networks and their modern variants, have garnered renewed interest due to advancements in memory capacity and connections with self-attention in transformers. In this work, we introduce a unified framework-Hopfield-Fenchel-Young networks-which generalizes these models to a broader family of energy functions. Our energies are formulated as the difference between two Fenchel-Young losses: one, parameterized by a generalized entropy, defines the Hopfield scoring mechanism, while the other applies a post-transformation to the Hopfield output. By utilizing Tsallis and norm entropies, we derive end-to-end differentiable update rules that enable sparse transformations, uncovering new connections between loss margins, sparsity, and exact retrieval of single memory patterns. We further extend this framework to structured Hopfield networks using the SparseMAP transformation, allowing the retrieval of pattern associations rather than a single pattern. Our framework unifies and extends traditional and modern Hopfield networks and provides an energy minimization perspective for widely used post-transformations like $\ell_2$-normalization and layer normalization-all through suitable choices of Fenchel-Young losses and by using convex analysis as a building block. Finally, we validate our Hopfield-Fenchel-Young networks on diverse memory recall tasks, including free and sequential recall. Experiments on simulated data, image retrieval, multiple instance learning, and text rationalization demonstrate the effectiveness of our approach., Comment: 49 pages, 14 figures. arXiv admin note: text overlap with arXiv:2402.13725

Driving the nuclear fusion reaction p+11B -> 3 alpha + 8.7 MeV in laboratory conditions, by interaction between high-power laser pulses and matter, has become a popular field of research, due to numerous applications that it can potentially allow: an alternative to deuterium-tritium (DT) for fusion energy production, astrophysics studies and alpha-particle generation for medical treatments. A possible scheme for laser-driven p-11B reactions is to direct a beam of laser-accelerated protons onto a boron sample (the so-called 'pitcher-catcher' scheme). This technique was successfully implemented on large, energetic lasers, yielding hundreds of joules per shot at low repetition. We present here a complementary approach, exploiting the high-repetition rate of the VEGA III petawatt laser at CLPU (Spain), aiming at accumulating results from many interactions at much lower energy, for better controlling the parameters and the statistics of the measurements. Despite a moderate energy per pulse, our experiment allowed exploring the laser-driven fusion process with tens (up to hundreds) of laser shots. The experiment provided a clear signature of the produced reactions and of the fusion products, accumulated over many shots, leading to an improved optimization of the diagnostic for these experimental campaigns In this paper we discuss the effectiveness of the laser-driven p-11B fusion in the pitcher-catcher scheme, at high-repetition rate, addressing the challenges of this experimental scheme and highlighting its critical aspects. Our proposed methodologies allow evaluating the performance of this scheme for laser-driven alpha particle production and can be adapted to high-repetition rate laser facilities with higher energy and intensity.

Extremum Seeking Control (ESC) is a well-known set of continuous time algorithms for model-free optimization of a cost function. One issue for ESCs is the convergence rates of parameters to extrema of unknown cost functions. The local convergence rate depends on the second, or sometimes higher, order derivatives of the unknown cost function. To mitigate this dependency, we propose the use of the RMSprop optimizer for ESCs as RMSprop is an adaptive gradient-based optimizer which attempts to have a normalized convergence rate in all parameters. Practical stability results are given for this RMSprop ESC (RMSpESC). In particular notability, the proof of practical stability uses Lyapunov function based on observed contracting, attractive sets. Versions of this Lyapunov function could be applied to other areas of applications, in particular for interconnected systems., Comment: 6 pages, 1 figure, L-CSS and American Control Conference Submission

For a map that is strictly but not strongly convex, model-based gradient extremum seeking has an eigenvalue of zero at the extremum, i.e., it fails at exponential convergence. Interestingly, perturbation-based model-free extremum seeking has a negative Jacobian, in the average, meaning that its (practical) convergence is exponential, even though the map's Hessian is zero at the extremum. While these observations for the gradient algorithm are not trivial, we focus in this paper on an even more nontrivial study of the same phenomenon for Newton-based extremum seeking control (NESC). NESC is a second-order method which corrects for the unknown Hessian of the unknown map, not only in order to speed up parameter convergence, but also (1) to make the convergence rate user-assignable in spite of the unknown Hessian, and (2) to equalize the convergence rates in different directions for multivariable maps. Previous NESC work established stability only for maps whose Hessians are strictly positive definite everywhere, so the Hessian is invertible everywhere. For a scalar map, we establish the rather unexpected property that, even when the map behind is strictly convex but not strongly convex, i.e., when the Hessian may be zero, NESC guarantees practical asymptotic stability, semiglobally. While a model-based Newton-based algorithm would run into non-invertibility of the Hessian, the perturbation-based NESC, surprisingly, avoids this challenge by leveraging the fact that the average of the perturbation-based Hessian estimate is always positive, even though the actual Hessian may be zero., Comment: 6 pages, 5 figures

Large Language Models (LLMs) have enabled new ways to satisfy information needs. Although great strides have been made in applying them to settings like document ranking and short-form text generation, they still struggle to compose complete, accurate, and verifiable long-form reports. Reports with these qualities are necessary to satisfy the complex, nuanced, or multi-faceted information needs of users. In this perspective paper, we draw together opinions from industry and academia, and from a variety of related research areas, to present our vision for automatic report generation, and -- critically -- a flexible framework by which such reports can be evaluated. In contrast with other summarization tasks, automatic report generation starts with a detailed description of an information need, stating the necessary background, requirements, and scope of the report. Further, the generated reports should be complete, accurate, and verifiable. These qualities, which are desirable -- if not required -- in many analytic report-writing settings, require rethinking how to build and evaluate systems that exhibit these qualities. To foster new efforts in building these systems, we present an evaluation framework that draws on ideas found in various evaluations. To test completeness and accuracy, the framework uses nuggets of information, expressed as questions and answers, that need to be part of any high-quality generated report. Additionally, evaluation of citations that map claims made in the report to their source documents ensures verifiability., Comment: 12 pages, 4 figures, accepted at SIGIR 2024 as perspective paper

This paper describes the submission runs from the HLTCOE team at the CIRAL CLIR tasks for African languages at FIRE 2023. Our submissions use machine translation models to translate the documents and the training passages, and ColBERT-X as the retrieval model. Additionally, we present a set of unofficial runs that use an alternative training procedure with a similar training setting., Comment: 10 pages, 2 figures. System description paper for HLTCOE's participation in CIRAL@FIRE 2023

The principal goal of the TREC Neural Cross-Language Information Retrieval (NeuCLIR) track is to study the impact of neural approaches to cross-language information retrieval. The track has created four collections, large collections of Chinese, Persian, and Russian newswire and a smaller collection of Chinese scientific abstracts. The principal tasks are ranked retrieval of news in one of the three languages, using English topics. Results for a multilingual task, also with English topics but with documents from all three newswire collections, are also reported. New in this second year of the track is a pilot technical documents CLIR task for ranked retrieval of Chinese technical documents using English topics. A total of 220 runs across all tasks were submitted by six participating teams and, as baselines, by track coordinators. Task descriptions and results are presented., Comment: 27 pages, 17 figures. Part of the TREC 2023 Proceedings

Modern Hopfield networks have enjoyed recent interest due to their connection to attention in transformers. Our paper provides a unified framework for sparse Hopfield networks by establishing a link with Fenchel-Young losses. The result is a new family of Hopfield-Fenchel-Young energies whose update rules are end-to-end differentiable sparse transformations. We reveal a connection between loss margins, sparsity, and exact memory retrieval. We further extend this framework to structured Hopfield networks via the SparseMAP transformation, which can retrieve pattern associations instead of a single pattern. Experiments on multiple instance learning and text rationalization demonstrate the usefulness of our approach., Comment: 20 pages, 4 figures

Laser technologies improved after the understanding of the Chirped Pulse Amplification (CPA) which allows energetic laser beams to be compressed to tens of femtosecond (fs) pulse durations and focused to few $\mu$m. Protons of tens of MeV can be accelerated using for instance the Target Normal Sheath Acceleration (TNSA) method and focused on secondary targets. In such conditions, nuclear reactions can occur and radioisotopes relevant for medical purposes be produced. High repetition lasers can be used to produce enough isotopes for medical applications. This route is competitive to conventional methods mostly based on accelerators. In this paper we study the production of $^{67}$Cu, $^{63}$Zn, $^{18}$F and $^{11}$C currently used in positron emission tomography (PET) and other applications. At the same time, we study the reaction $^{10}$B(p,$\alpha$)$^{7}$Be and $^{70}$Zn(p,4n)$^{67}$Ga to put further constraints to the proton distributions at different angles and to the reaction $^{11}$B(p,$\alpha$)$^{8}$Be relevant for energy production. The experiment was performed at the 1 petawatt (PW) laser facility at Vega III located in Salamanca-Spain. Angular distributions of radioisotopes in the forward (with respect to the laser direction) and backward directions were measured using a High Purity Germanium Detector (HPGE). Our results are reasonably reproduced by the numerical estimates following the approach of Kimura et al. (NIMA637(2011)167)

This study addresses the application of deep learning techniques in joint sound signal classification and localization networks. Current state-of-the-art sound source localization deep learning networks lack feature aggregation within their architecture. Feature aggregation enhances model performance by enabling the consolidation of information from different feature scales, thereby improving feature robustness and invariance. This is particularly important in SSL networks, which must differentiate direct and indirect acoustic signals. To address this gap, we adapt feature aggregation techniques from computer vision neural networks to signal detection neural networks. Additionally, we propose the Scale Encoding Network (SEN) for feature aggregation to encode features from various scales, compressing the network for more computationally efficient aggregation. To evaluate the efficacy of feature aggregation in SSL networks, we integrated the following computer vision feature aggregation sub-architectures into a SSL control architecture: Path Aggregation Network (PANet), Weighted Bi-directional Feature Pyramid Network (BiFPN), and SEN. These sub-architectures were evaluated using two metrics for signal classification and two metrics for direction-of-arrival regression. PANet and BiFPN are established aggregators in computer vision models, while the proposed SEN is a more compact aggregator. The results suggest that models incorporating feature aggregations outperformed the control model, the Sound Event Localization and Detection network (SELDnet), in both sound signal classification and localization. The feature aggregation techniques enhance the performance of sound detection neural networks, particularly in direction-of-arrival regression.

In recent years, advancements have been made towards the goal of using chaotic coverage path planners for autonomous search and traversal of spaces with limited environmental cues. However, the state of this field is still in its infancy as there has been little experimental work done. The existing experimental works have not developed robust methods to satisfactorily address the immediate set of problems a chaotic coverage path planner needs to overcome in order to scan realistic environments within reasonable coverage times. These immediate problems are as follows: (1) an obstacle avoidance technique that reduces halts or disruptions in continuous chaotic trajectories, (2) a means to spread chaotic trajectories across the environment (especially crucial for large and/or complex-shaped environments) that need to be covered, and (3) a real-time coverage calculation technique that is accurate and independent of cell size. This study addresses these problems by developing a novel applied framework for real-world applications of chaotic coverage path planners while providing techniques for effective obstacle avoidance, chaotic trajectory dispersal, and accurate real-time coverage calculation. These algorithms were created within the ROS framework and make up a newly developed chaotic path planning application. The performance of this application was comparable to that of a conventional optimal path planner. The performance tests were carried out in environments of various sizes, shapes, and obstacle densities, both in real-life and Gazebo simulations.

This is the first year of the TREC Neural CLIR (NeuCLIR) track, which aims to study the impact of neural approaches to cross-language information retrieval. The main task in this year's track was ad hoc ranked retrieval of Chinese, Persian, or Russian newswire documents using queries expressed in English. Topics were developed using standard TREC processes, except that topics developed by an annotator for one language were assessed by a different annotator when evaluating that topic on a different language. There were 172 total runs submitted by twelve teams., Comment: 22 pages, 13 figures, 10 tables. Part of the Thirty-First Text REtrieval Conference (TREC 2022) Proceedings. Replace the misplaced Russian result table

Studying systems where many individual bodies in motion interact with one another is a complex and interesting area. Simple mechanisms that may be determined for biological, chemical, or physical reasons can lead to astonishingly complex results that require a further understanding of the moving bodies. With the increasing interaction between computation and various scientific areas, it has become more useful, feasible, and important to create models for these systems. Here, we present two families of models, deterministic and nondeterministic, along with three distinct and realistic interaction mechanisms. These are combined in a unique way to provide the groundwork for particle system models across multiple disciplines. This work has applications that range from biology to chemistry and physics. In addition to describing the motivations and math behind all the models, software is provided that allows researchers to quickly adjust and implement what is described here., Comment: 11 pages, 1 figure

The advent of transformer-based models such as BERT has led to the rise of neural ranking models. These models have improved the effectiveness of retrieval systems well beyond that of lexical term matching models such as BM25. While monolingual retrieval tasks have benefited from large-scale training collections such as MS MARCO and advances in neural architectures, cross-language retrieval tasks have fallen behind these advancements. This paper introduces ColBERT-X, a generalization of the ColBERT multi-representation dense retrieval model that uses the XLM-RoBERTa (XLM-R) encoder to support cross-language information retrieval (CLIR). ColBERT-X can be trained in two ways. In zero-shot training, the system is trained on the English MS MARCO collection, relying on the XLM-R encoder for cross-language mappings. In translate-train, the system is trained on the MS MARCO English queries coupled with machine translations of the associated MS MARCO passages. Results on ad hoc document ranking tasks in several languages demonstrate substantial and statistically significant improvements of these trained dense retrieval models over traditional lexical CLIR baselines., Comment: Accepted at ECIR 2022 (Full paper)

Policies trained via reinforcement learning (RL) are often very complex even for simple tasks. In an episode with $n$ time steps, a policy will make $n$ decisions on actions to take, many of which may appear non-intuitive to the observer. Moreover, it is not clear which of these decisions directly contribute towards achieving the reward and how significant is their contribution. Given a trained policy, we propose a black-box method based on counterfactual reasoning that estimates the causal effect that these decisions have on reward attainment and ranks the decisions according to this estimate. In this preliminary work, we compare our measure against an alternative, non-causal, ranking procedure, highlight the benefits of causality-based policy ranking, and discuss potential future work integrating causal algorithms into the interpretation of RL agent policies., Comment: preprint; 6 pages. arXiv admin note: substantial text overlap with arXiv:2008.13607

The ability to adapt to changes in environmental contingencies is an important challenge in reinforcement learning. Indeed, transferring previously acquired knowledge to environments with unseen structural properties can greatly enhance the flexibility and efficiency by which novel optimal policies may be constructed. In this work, we study the problem of transfer learning under changes in the environment dynamics. In this study, we apply causal reasoning in the offline reinforcement learning setting to transfer a learned policy to new environments. Specifically, we use the Decision Transformer (DT) architecture to distill a new policy on the new environment. The DT is trained on data collected by performing policy rollouts on factual and counterfactual simulations from the source environment. We show that this mechanism can bootstrap a successful policy on the target environment while retaining most of the reward.

Responsive poly-N-isopropylacrylamide-based microgels are commonly used as model colloids with soft repulsive interactions. It has been shown that the microgel-microgel interaction in solution can be easily adjusted by varying the environmental parameters, e.g., temperature, pH, or salt concentration. Furthermore, microgels readily adsorb to liquid-gas and liquid-liquid interfaces forming responsive foams and emulsions that can be broken on-demand. In this work, we explore the interactions between microgel monolayers at the air-water interface and a hard colloid in the water. Force-distance curves between the monolayer and a silica particle were measured with the Monolayer Particle Interaction Apparatus. The measurements were conducted at different temperatures and lateral compression, i.e., different surface pressures. The force-distance approach curves display long-range repulsive forces below the volume phase transition temperature of the microgels. Temperature and lateral compression reduce the stiffness of the monolayer. The adhesion increases with temperature and decreases with a lateral compression of the monolayer. When compressed laterally, the interactions between the microgels are hardly affected by temperature, as the directly adsorbed microgel fractions are nearly insensitive to temperature. In contrast, our findings show that the temperature-dependent swelling of the microgel fractions in the aqueous phase strongly influences the interaction with the probe. The microgel monolayer changes from a soft to a hard repulsive interface.

Surveillance and exploration of large environments is a tedious task. In spaces with limited environmental cues, random-like search is an effective approach as it allows the robot to perform online coverage of environments using simple algorithm designs. One way to generate random-like scanning search is to use nonlinear dynamical systems to impart chaos into the searching robot's controller. This will result in the generation of unpredictable yet deterministic trajectories, allowing designers to control the system and achieve a high scanning coverage of an area. However, the unpredictability comes at the cost of increased coverage time and a lack of scalability, both of which have been ignored by the state-of-the-art chaotic path planners. This work introduces a new, scalable technique that helps a robot to steer away from the obstacles and cover the entire search space in a short period of time. The technique involves coupling and manipulating two chaotic systems to reduce the coverage time and enable scanning of unknown environments with different online properties. Using this new technique resulted in an average 49% boost in the robot's performance compared to the state-of-the-art planners. the overall search performance of the chaotic planner remained comparable to optimal systems while still ensuring unpredictable paths.

Edge computing offers an additional layer of compute infrastructure closer to the data source before raw data from privacy-sensitive and performance-critical applications is transferred to a cloud data center. Deep Neural Networks (DNNs) are one class of applications that are reported to benefit from collaboratively computing between the edge and the cloud. A DNN is partitioned such that specific layers of the DNN are deployed onto the edge and the cloud to meet performance and privacy objectives. However, there is limited understanding of: (a) whether and how evolving operational conditions (increased CPU and memory utilization at the edge or reduced data transfer rates between the edge and the cloud) affect the performance of already deployed DNNs, and (b) whether a new partition configuration is required to maximize performance. A DNN that adapts to changing operational conditions is referred to as an 'adaptive DNN'. This paper investigates whether there is a case for adaptive DNNs in edge computing by considering three questions: (i) Are DNNs sensitive to operational conditions? (ii) How sensitive are DNNs to operational conditions? (iii) Do individual or a combination of operational conditions equally affect DNNs? (iv) Is DNN partitioning sensitive to hardware architectures on the cloud/edge? The exploration is carried out in the context of 8 pre-trained DNN models and the results presented are from analyzing nearly 8 million data points. The results highlight that network conditions affects DNN performance more than CPU or memory related operational conditions. Repartitioning is noted to provide a performance gain in a number of cases, but a specific trend was not noted in relation to its correlation to the underlying hardware architecture. Nonetheless, the need for adaptive DNNs is confirmed.

We develop a normative framework for hierarchical model-based policy optimization based on applying second-order methods in the space of all possible state-action paths. The resulting natural path gradient performs policy updates in a manner which is sensitive to the long-range correlational structure of the induced stationary state-action densities. We demonstrate that the natural path gradient can be computed exactly given an environment dynamics model and depends on expressions akin to higher-order successor representations. In simulation, we show that the priorization of local policy updates in the resulting policy flow indeed reflects the intuitive state-space hierarchy in several toy problems., Comment: NeurIPS 2019 Optimization Foundations of Reinforcement Learning Workshop. v2: typos fixed, minor edits for improved clarity

We introduce a curriculum learning approach to adapt generic neural machine translation models to a specific domain. Samples are grouped by their similarities to the domain of interest and each group is fed to the training algorithm with a particular schedule. This approach is simple to implement on top of any neural framework or architecture, and consistently outperforms both unadapted and adapted baselines in experiments with two distinct domains and two language pairs.

Machine translation systems based on deep neural networks are expensive to train. Curriculum learning aims to address this issue by choosing the order in which samples are presented during training to help train better models faster. We adopt a probabilistic view of curriculum learning, which lets us flexibly evaluate the impact of curricula design, and perform an extensive exploration on a German-English translation task. Results show that it is possible to improve convergence time at no loss in translation quality. However, results are highly sensitive to the choice of sample difficulty criteria, curriculum schedule and other hyperparameters.

To better understand the effectiveness of continued training, we analyze the major components of a neural machine translation system (the encoder, decoder, and each embedding space) and consider each component's contribution to, and capacity for, domain adaptation. We find that freezing any single component during continued training has minimal impact on performance, and that performance is surprisingly good when a single component is adapted while holding the rest of the model fixed. We also find that continued training does not move the model very far from the out-of-domain model, compared to a sensitivity analysis metric, suggesting that the out-of-domain model can provide a good generic initialization for the new domain., Comment: presented at WMT 2018. Please cite using the bib entry from here: http://www.statmt.org/wmt18/bib/WMT013.bib

Hierarchies are of fundamental interest in both stochastic optimal control and biological control due to their facilitation of a range of desirable computational traits in a control algorithm and the possibility that they may form a core principle of sensorimotor and cognitive control systems. However, a theoretically justified construction of state-space hierarchies over all spatial resolutions and their evolution through a policy inference process remains elusive. Here, a formalism for deriving such normative representations of discrete Markov decision processes is introduced in the context of graphs. The resulting hierarchies correspond to a hierarchical policy inference algorithm approximating a discrete gradient flow between state-space trajectory densities generated by the prior and optimal policies., Comment: NIPS 2017 Workshop on Hierarchical Reinforcement Learning. 8 pages, 1 figure

The paper summarizes the development of the LVCSR system built as a part of the Pashto speech-translation system at the SCALE (Summer Camp for Applied Language Exploration) 2015 workshop on "Speech-to-text-translation for low-resource languages". The Pashto language was chosen as a good "proxy" low-resource language, exhibiting multiple phenomena which make the speech-recognition and and speech-to-text-translation systems development hard. Even when the amount of data is seemingly sufficient, given the fact that the data originates from multiple sources, the preliminary experiments reveal that there is little to no benefit in merging (concatenating) the corpora and more elaborate ways of making use of all of the data must be worked out. This paper concentrates only on the LVCSR part and presents a range of different techniques that were found to be useful in order to benefit from multiple different corpora