Showing 1–15 of 15 results for author: Jacob, F

An Application of Large Language Models to Coding Negotiation Transcripts

Authors: Ray Friedman, Jaewoo Cho, Jeanne Brett, Xuhui Zhan, Ningyu Han, Sriram Kannan, Yingxiang Ma, Jesse Spencer-Smith, Elisabeth Jäckel, Alfred Zerres, Madison Hooper, Katie Babbit, Manish Acharya, Wendi Adair, Soroush Aslani, Tayfun Aykaç, Chris Bauman, Rebecca Bennett, Garrett Brady, Peggy Briggs, Cheryl Dowie, Chase Eck, Igmar Geiger, Frank Jacob, Molly Kern , et al. (33 additional authors not shown)

Abstract: In recent years, Large Language Models (LLM) have demonstrated impressive capabilities in the field of natural language processing (NLP). This paper explores the application of LLMs in negotiation transcript analysis by the Vanderbilt AI Negotiation Lab. Starting in September 2022, we applied multiple strategies using LLMs from zero shot learning to fine tuning models to in-context learning). The… ▽ More In recent years, Large Language Models (LLM) have demonstrated impressive capabilities in the field of natural language processing (NLP). This paper explores the application of LLMs in negotiation transcript analysis by the Vanderbilt AI Negotiation Lab. Starting in September 2022, we applied multiple strategies using LLMs from zero shot learning to fine tuning models to in-context learning). The final strategy we developed is explained, along with how to access and use the model. This study provides a sense of both the opportunities and roadblocks for the implementation of LLMs in real life applications and offers a model for how LLMs can be applied to coding in other fields. △ Less

Submitted 18 July, 2024; originally announced July 2024.

How to Raise a Robot -- A Case for Neuro-Symbolic AI in Constrained Task Planning for Humanoid Assistive Robots

Authors: Niklas Hemken, Florian Jacob, Fabian Peller-Konrad, Rainer Kartmann, Tamim Asfour, Hannes Hartenstein

Abstract: Humanoid robots will be able to assist humans in their daily life, in particular due to their versatile action capabilities. However, while these robots need a certain degree of autonomy to learn and explore, they also should respect various constraints, for access control and beyond. We explore the novel field of incorporating privacy, security, and access control constraints with robot task plan… ▽ More Humanoid robots will be able to assist humans in their daily life, in particular due to their versatile action capabilities. However, while these robots need a certain degree of autonomy to learn and explore, they also should respect various constraints, for access control and beyond. We explore the novel field of incorporating privacy, security, and access control constraints with robot task planning approaches. We report preliminary results on the classical symbolic approach, deep-learned neural networks, and modern ideas using large language models as knowledge base. From analyzing their trade-offs, we conclude that a hybrid approach is necessary, and thereby present a new use case for the emerging field of neuro-symbolic artificial intelligence. △ Less

Submitted 27 December, 2023; v1 submitted 14 December, 2023; originally announced December 2023.

Comments: 8 pages, follow-up extended version of our SACMAT 2023 poster abstract: "Poster: How to Raise a Robot - Beyond Access Control Constraints in Assistive Humanoid Robots" https://dl.acm.org/doi/abs/10.1145/3589608.3595078

On Extend-Only Directed Posets and Derived Byzantine-Tolerant Replicated Data Types (Extended Version)

Authors: Florian Jacob, Hannes Hartenstein

Abstract: We uncover the extend-only directed posets (EDP) structure as a unification of recently discussed DAG-based Byzantine-tolerant conflict-free replicated data types (CRDT). We also show how a key-value map model can be derived from the EDP formulation, and give an outlook on an EDP-based systemic access control CRDT as a formalization of the CRDT used in the Matrix messaging system. We uncover the extend-only directed posets (EDP) structure as a unification of recently discussed DAG-based Byzantine-tolerant conflict-free replicated data types (CRDT). We also show how a key-value map model can be derived from the EDP formulation, and give an outlook on an EDP-based systemic access control CRDT as a formalization of the CRDT used in the Matrix messaging system. △ Less

Submitted 9 April, 2023; originally announced April 2023.

Comments: With the inclusion of an appendix of a formalization and CRDT proof sketch of an EDP-based CRDT with systemic access control, this is an extended version of the paper presented at the 10th Workshop on Principles and Practice of Consistency for Distributed Data (PaPoC), 2023-05-08, Rome, Italy

Radio observations of the Black Hole X-ray Binary EXO 1846-031 re-awakening from a 34-year slumber

Authors: D. R. A. Williams, S. E. Motta, R. Fender, J. C. A. Miller-Jones, J. Neilsen, J. R. Allison, J. Bright, I. Heywood, P. F. L. Jacob, L. Rhodes, E. Tremou, P. Woudt, J. van den Eijnden, F. Carotenuto, D. A. Green, D. Titterington, A. J. van der Horst, P. Saikia

Abstract: We present radio [1.3 GHz MeerKAT, 4-8 GHz Karl G. Jansky Very Large Array (VLA) and 15.5 GHz Arcminute Microkelvin Imager Large Array (AMI-LA)] and X-ray (Swift and MAXI) data from the 2019 outburst of the candidate Black Hole X-ray Binary (BHXB) EXO 1846-031. We compute a Hardness-Intensity diagram, which shows the characteristic q-shaped hysteresis of BHXBs in outburst. EXO 1846-031 was monitor… ▽ More We present radio [1.3 GHz MeerKAT, 4-8 GHz Karl G. Jansky Very Large Array (VLA) and 15.5 GHz Arcminute Microkelvin Imager Large Array (AMI-LA)] and X-ray (Swift and MAXI) data from the 2019 outburst of the candidate Black Hole X-ray Binary (BHXB) EXO 1846-031. We compute a Hardness-Intensity diagram, which shows the characteristic q-shaped hysteresis of BHXBs in outburst. EXO 1846-031 was monitored weekly with MeerKAT and approximately daily with AMI-LA. The VLA observations provide sub-arcsecond-resolution images at key points in the outburst, showing moving radio components. The radio and X-ray light curves broadly follow each other, showing a peak on ~MJD 58702, followed by a short decline before a second peak between ~MJD 58731-58739. We estimate the minimum energy of these radio flares from equipartition, calculating values of $E_{\rm min} \sim$ 4$\times$10$^{41}$ and 5$\times$10$^{42}$ erg, respectively. The exact date of the return to `quiescence' is missed in the X-ray and radio observations, but we suggest that it likely occurred between MJD 58887 and 58905. From the Swift X-ray flux on MJD 58905 and assuming the soft-to-hard transition happened at 0.3-3 per cent Eddington, we calculate a distance range of 2.4-7.5\,kpc. We computed the radio:X-ray plane for EXO 1846-031 in the `hard' state, showing that it is most likely a `radio-quiet' BH, preferentially at 4.5 kpc. Using this distance and a jet inclination angle of $θ$=73$^{\circ}$, the VLA data place limits on the intrinsic jet speed of $β_{\rm int} = 0.29c$, indicating sub-luminal jet motion. △ Less

Submitted 21 September, 2022; originally announced September 2022.

Comments: Accepted for publication in MNRAS on 20 September 2022, 17 pages, 6 figures

On Conflict-Free Replicated Data Types and Equivocation in Byzantine Setups

Authors: Florian Jacob, Saskia Bayreuther, Hannes Hartenstein

Abstract: We explore the property of equivocation tolerance for Conflict-Free Replicated Data Types (CRDTs). We show that a subclass of CRDTs is equivocation-tolerant and can thereby cope with any number of Byzantine faults: Without equivocation detection, prevention or remediation, they still fulfill strong eventual consistency (SEC). We also conjecture that there is only one operation-based CRDT design su… ▽ More We explore the property of equivocation tolerance for Conflict-Free Replicated Data Types (CRDTs). We show that a subclass of CRDTs is equivocation-tolerant and can thereby cope with any number of Byzantine faults: Without equivocation detection, prevention or remediation, they still fulfill strong eventual consistency (SEC). We also conjecture that there is only one operation-based CRDT design supporting non-commutative operations that fulfills SEC in Byzantine environments with any number of faults. △ Less

Submitted 8 October, 2021; v1 submitted 22 September, 2021; originally announced September 2021.

Analysis of the Matrix Event Graph Replicated Data Type

Authors: Florian Jacob, Carolin Beer, Norbert Henze, Hannes Hartenstein

Abstract: Matrix is a new kind of decentralized, topic-based publish-subscribe middleware for communication and data storage that is getting popular particularly as a basis for secure instant messaging. In comparison to traditional decentralized communication systems, Matrix replaces pure message passing with a replicated data structure. This data structure, which we extract and call the Matrix Event Graph… ▽ More Matrix is a new kind of decentralized, topic-based publish-subscribe middleware for communication and data storage that is getting popular particularly as a basis for secure instant messaging. In comparison to traditional decentralized communication systems, Matrix replaces pure message passing with a replicated data structure. This data structure, which we extract and call the Matrix Event Graph (MEG), depicts the causal history of messages. We show that this MEG represents an interesting and important replicated data type for general decentralized applications that are based on causal histories of publish-subscribe events: we show that a MEG possesses strong properties with respect to consistency, byzantine attackers, and scalability. First, we show that the MEG provides Strong Eventual Consistency (SEC), and that it is available under partition, by proving that the MEG is a Conflict-Free Replicated Data Type for causal histories. While strong consistency is impossible here as shown by the famous CAP theorem, SEC is among the best known achievable trade-offs. Second, we discuss the implications of byzantine attackers on the data type's properties. We note that the MEG, as it does not strive for consensus, can cope with $n > f$ environments with $n$ total participants of which $f$ show byzantine faults. Furthermore, we analyze scalability: Using Markov chains we study the width of the MEG, defined as the number of forward extremities, over time and observe an almost optimal evolution. We conjecture that this property is inherent to the underlying spatially inhomogeneous random walk. △ Less

Submitted 12 November, 2020; originally announced November 2020.

Comments: 14 pages, 5 figures

MSC Class: 60J20 ACM Class: E.1; C.2.4; G.3

Measuring the distance to the black hole candidate X-ray binary MAXI J1348-630 using HI absorption

Authors: J. Chauhan, J. C. A. Miller-Jones, W. Raja, J. R. Allison, P. F. L. Jacob, G. E. Anderson, F. Carotenuto, S. Corbel, R. Fender, A. Hotan, M. Whiting, P. A. Woudt, B. Koribalski, E. Mahony

Abstract: We present HI absorption spectra of the black hole candidate X-ray binary (XRB) MAXI J1348-630 using the Australian Square Kilometre Array Pathfinder (ASKAP) and MeerKAT. The ASKAP HI spectrum shows a maximum negative radial velocity (with respect to the local standard of rest) of $-31\pm4$ km s$^{-1}$ for MAXI J1348-630, as compared to $-50\pm4$ km s$^{-1}$ for a stacked spectrum of several nearb… ▽ More We present HI absorption spectra of the black hole candidate X-ray binary (XRB) MAXI J1348-630 using the Australian Square Kilometre Array Pathfinder (ASKAP) and MeerKAT. The ASKAP HI spectrum shows a maximum negative radial velocity (with respect to the local standard of rest) of $-31\pm4$ km s$^{-1}$ for MAXI J1348-630, as compared to $-50\pm4$ km s$^{-1}$ for a stacked spectrum of several nearby extragalactic sources. This implies a most probable distance of $2.2^{+0.5}_{-0.6}$ kpc for MAXI J1348-630, and a strong upper limit of the tangent point distance at $5.3\pm0.1$ kpc. Our preferred distance implies that MAXI J1348-630 reached $17\pm10$ % of the Eddington luminosity at the peak of its outburst, and that the source transited from the soft to the hard X-ray spectral state at $2.5\pm1.5$ % of the Eddington luminosity. The MeerKAT HI spectrum of MAXI J1348-630 (obtained from the older, low-resolution 4k mode) is consistent with the re-binned ASKAP spectrum, highlighting the potential of the eventual capabilities of MeerKAT for XRB spectral line studies. △ Less

Submitted 4 December, 2020; v1 submitted 30 September, 2020; originally announced September 2020.

Comments: Accepted for publication in MNRAS Letters

A Glimpse of the Matrix (Extended Version): Scalability Issues of a New Message-Oriented Data Synchronization Middleware

Authors: Florian Jacob, Jan Grashöfer, Hannes Hartenstein

Abstract: Matrix is a new message-oriented data synchronization middleware, used as a federated platform for near real-time decentralized applications. It features a novel approach for inter-server communication based on synchronizing message history by using a replicated data structure. We measured the structure of public parts in the Matrix federation as a basis to analyze the middleware's scalability. We… ▽ More Matrix is a new message-oriented data synchronization middleware, used as a federated platform for near real-time decentralized applications. It features a novel approach for inter-server communication based on synchronizing message history by using a replicated data structure. We measured the structure of public parts in the Matrix federation as a basis to analyze the middleware's scalability. We confirm that users are currently cumulated on a single large server, but find more small servers than expected. We then analyze network load distribution in the measured structure and identify scalability issues of Matrix' group communication mechanism in structurally diverse federations. △ Less

Submitted 29 November, 2019; v1 submitted 14 October, 2019; originally announced October 2019.

Comments: Extended tech report of the Poster Abstract https://doi.org/10.1145/3366627.3368106 from Middleware 2019

Multi-scale GANs for Memory-efficient Generation of High Resolution Medical Images

Authors: Hristina Uzunova, Jan Ehrhardt, Fabian Jacob, Alex Frydrychowicz, Heinz Handels

Abstract: Currently generative adversarial networks (GANs) are rarely applied to medical images of large sizes, especially 3D volumes, due to their large computational demand. We propose a novel multi-scale patch-based GAN approach to generate large high resolution 2D and 3D images. Our key idea is to first learn a low-resolution version of the image and then generate patches of successively growing resolut… ▽ More Currently generative adversarial networks (GANs) are rarely applied to medical images of large sizes, especially 3D volumes, due to their large computational demand. We propose a novel multi-scale patch-based GAN approach to generate large high resolution 2D and 3D images. Our key idea is to first learn a low-resolution version of the image and then generate patches of successively growing resolutions conditioned on previous scales. In a domain translation use-case scenario, 3D thorax CTs of size 512x512x512 and thorax X-rays of size 2048x2048 are generated and we show that, due to the constant GPU memory demand of our method, arbitrarily large images of high resolution can be generated. Moreover, compared to common patch-based approaches, our multi-resolution scheme enables better image quality and prevents patch artifacts. △ Less

Submitted 8 July, 2019; v1 submitted 2 July, 2019; originally announced July 2019.

Comments: Accepted at MICCAI 2019

Adoção de Social CRM em Micro e Pequenas Empresas: Uma Análise do Mercado Santareno

Authors: Gustavo Nogueira de Sousa, Luan Vinícius Huppes, Antônio Fernando Lavareda Jacob Jr, Fábio Manoel França Lobato

Abstract: Online social networks have changed the ways of communication and social interactions, especially in the Customer Relationship Management (CRM). In this sense, a new concept about business strategies involving CRM and social media has aroused, known as Social Customer Relationship Management. Despite to be an emergent and promising research field, it was perceived that Micro and Small Enterprises… ▽ More Online social networks have changed the ways of communication and social interactions, especially in the Customer Relationship Management (CRM). In this sense, a new concept about business strategies involving CRM and social media has aroused, known as Social Customer Relationship Management. Despite to be an emergent and promising research field, it was perceived that Micro and Small Enterprises (MSE) have shown few or no process of Social CRM implemented. Aiming to test this hypothesis, this work conducts a market analysis in Santarém City, located in the Pará State, evaluating the adoption of Social CRM by MSE. The main contribution of this study is related to the understanding of the dynamics between Social CRM and MSE. As results, the construction of insights' list of products and solutions suitable for the implementation of Social CRM by MSE, with the potential to guide research and development projects in this area. △ Less

Submitted 26 November, 2018; originally announced November 2018.

Comments: in Portuguese, Paper presented at the 15th International Conference On Information Systems & Technology Management

Development of a Social Network for Research Support and Individual Well-being Improvement

Authors: Lucas V. A. Caldas, Antonio F. L. Jacob Jr., Simone S. C. Silva, Fernando A. R. Pontes, Fábio M. F. Lobato

Abstract: The ways of communication and social interactions are changing. Web users are becoming increasingly engaged with Online Social Networks (OSN), which has a significant impact on the relationship mechanisms between individuals and communities. Most OSN platforms have strict policies regarding data access, harming its usage in psychological and social phenomena studies, It is also impacting the devel… ▽ More The ways of communication and social interactions are changing. Web users are becoming increasingly engaged with Online Social Networks (OSN), which has a significant impact on the relationship mechanisms between individuals and communities. Most OSN platforms have strict policies regarding data access, harming its usage in psychological and social phenomena studies, It is also impacting the development of computational methods to evaluate and improve social and individual well-being via the web. Aiming to fill this gap, we propose a platform that brings together social networks dynamics with forum features, altogether with gamification elements, targeting researchers interested in obtaining access to user's data to study psychological and social phenomena. △ Less

Submitted 9 September, 2018; originally announced September 2018.

Comments: This paper was accepted in the IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining 2018

Segmentation of Scanning Tunneling Microscopy Images Using Variational Methods and Empirical Wavelets

Authors: Bui Kevin, Fauman Jacob, Kes David, Torres Mandiola Leticia, Ciomaga Adina, Salazar Ricardo, Bertozzi L. Andrea, Gilles Jerome, Guttentag I. Andrew, Weiss S. Paul

Abstract: In the fields of nanoscience and nanotechnology, it is important to be able to functionalize surfaces chemically for a wide variety of applications. Scanning tunneling microscopes (STMs) are important instruments in this area used to measure the surface structure and chemistry with better than molecular resolution. Self-assembly is frequently used to create monolayers that redefine the surface che… ▽ More In the fields of nanoscience and nanotechnology, it is important to be able to functionalize surfaces chemically for a wide variety of applications. Scanning tunneling microscopes (STMs) are important instruments in this area used to measure the surface structure and chemistry with better than molecular resolution. Self-assembly is frequently used to create monolayers that redefine the surface chemistry in just a single-molecule-thick layer. Indeed, STM images reveal rich information about the structure of self-assembled monolayers since they convey chemical and physical properties of the studied material. In order to assist in and to enhance the analysis of STM and other images, we propose and demonstrate an image-processing framework that produces two image segmentations: one is based on intensities (apparent heights in STM images) and the other is based on textural patterns. The proposed framework begins with a cartoon+texture decomposition, which separates an image into its cartoon and texture components. Afterward, the cartoon image is segmented by a modified multiphase version of the local Chan-Vese model, while the texture image is segmented by a combination of 2D empirical wavelet transform and a clustering algorithm. Overall, our proposed framework contains several new features, specifically in presenting a new application of cartoon+texture decomposition and of the empirical wavelet transforms and in developing a specialized framework to segment STM images and other data. To demonstrate the potential of our approach, we apply it to actual STM images of cyanide monolayers on Au\{111\} and present their corresponding segmentation results. △ Less

Submitted 24 April, 2018; originally announced April 2018.

Towards a Job Title Classification System

Authors: Faizan Javed, Matt McNair, Ferosh Jacob, Meng Zhao

Abstract: Document classification for text, images and other applicable entities has long been a focus of research in academia and also finds application in many industrial settings. Amidst a plethora of approaches to solve such problems, machine-learning techniques have found success in a variety of scenarios. In this paper we discuss the design of a machine learning-based semi-supervised job title classif… ▽ More Document classification for text, images and other applicable entities has long been a focus of research in academia and also finds application in many industrial settings. Amidst a plethora of approaches to solve such problems, machine-learning techniques have found success in a variety of scenarios. In this paper we discuss the design of a machine learning-based semi-supervised job title classification system for the online job recruitment domain currently in production at CareerBuilder.com and propose enhancements to it. The system leverages a varied collection of classification as well clustering algorithms. These algorithms are encompassed in an architecture that facilitates leveraging existing off-the-shelf machine learning tools and techniques while keeping into consideration the challenges of constructing a scalable classification system for a large taxonomy of categories. As a continuously evolving system that is still under development we first discuss the existing semi-supervised classification system which is composed of both clustering and classification components in a proximity-based classifier setup and results of which are already used across numerous products at CareerBuilder. We then elucidate our long-term goals for job title classification and propose enhancements to the existing system in the form of a two-stage coarse and fine level classifier augmentation to construct a cascade of hierarchical vertical classifiers. Preliminary results are presented using experimental evaluation on real world industrial data. △ Less

Submitted 2 June, 2016; originally announced June 2016.

Absolute Geometry Calibration of Distributed Microphone Arrays in an Audio-Visual Sensor Network

Authors: Florian Jacob, Reinhold Haeb-Umbach

Abstract: Joint audio-visual speaker tracking requires that the locations of microphones and cameras are known and that they are given in a common coordinate system. Sensor self-localization algorithms, however, are usually separately developed for either the acoustic or the visual modality and return their positions in a modality specific coordinate system, often with an unknown rotation, scaling and trans… ▽ More Joint audio-visual speaker tracking requires that the locations of microphones and cameras are known and that they are given in a common coordinate system. Sensor self-localization algorithms, however, are usually separately developed for either the acoustic or the visual modality and return their positions in a modality specific coordinate system, often with an unknown rotation, scaling and translation between the two. In this paper we propose two techniques to determine the positions of acoustic sensors in a common coordinate system, based on audio-visual correlates, i.e., events that are localized by both, microphones and cameras separately. The first approach maps the output of an acoustic self-calibration algorithm by estimating rotation, scale and translation to the visual coordinate system, while the second solves a joint system of equations with acoustic and visual directions of arrival as input. The evaluation of the two strategies reveals that joint calibration outperforms the mapping approach and achieves an overall calibration error of 0.20m even in reverberant environments. △ Less

Submitted 13 April, 2015; originally announced April 2015.

A GPU-accelerated Direct-sum Boundary Integral Poisson-Boltzmann Solver

Authors: Weihua Geng, Ferosh Jacob

Abstract: In this paper, we present a GPU-accelerated direct-sum boundary integral method to solve the linear Poisson-Boltzmann (PB) equation. In our method, a well-posed boundary integral formulation is used to ensure the fast convergence of Krylov subspace based linear algebraic solver such as the GMRES. The molecular surfaces are discretized with flat triangles and centroid collocation. To speed up our m… ▽ More In this paper, we present a GPU-accelerated direct-sum boundary integral method to solve the linear Poisson-Boltzmann (PB) equation. In our method, a well-posed boundary integral formulation is used to ensure the fast convergence of Krylov subspace based linear algebraic solver such as the GMRES. The molecular surfaces are discretized with flat triangles and centroid collocation. To speed up our method, we take advantage of the parallel nature of the boundary integral formulation and parallelize the schemes within CUDA shared memory architecture on GPU. The schemes use only $11N+6N_c$ size-of-double device memory for a biomolecule with $N$ triangular surface elements and $N_c$ partial charges. Numerical tests of these schemes show well-maintained accuracy and fast convergence. The GPU implementation using one GPU card (Nvidia Tesla M2070) achieves 120-150X speed-up to the implementation using one CPU (Intel L5640 2.27GHz). With our approach, solving PB equations on well-discretized molecular surfaces with up to 300,000 boundary elements will take less than about 10 minutes, hence our approach is particularly suitable for fast electrostatics computations on small to medium biomolecules. △ Less

Submitted 24 January, 2013; originally announced January 2013.