Papers by Yasser Mohammad
Autonomous agents engaging in automatic negotiations on behalf of humans or institutions are usua... more Autonomous agents engaging in automatic negotiations on behalf of humans or institutions are usually assumed to have full knowledge of the utility function for the actors they represent. In many cases, these utility functions are difficult to know apriori for every possible outcome of the negotiation. Moreover, it may not be necessary for the agent to know the utility of outcomes that are never offered or considered during the negotiation. State-of-the-art approaches to utility elicitation during negotiation assume that the agent can ask questions from a predefined countable set to reduce its uncertainty about the utility function. This paper extends that body of work by lifting the countability assumption providing an optimal algorithm for selecting the best outcome and utility level about which to ask the actor. The paper reports the results of comparing the proposed algorithm with state-of-the-art algorithms using both synthetic and realistic negotiation scenarios. These evaluati...
Despite abundant negotiation strategies in literature, the complexity of automated negotiation fo... more Despite abundant negotiation strategies in literature, the complexity of automated negotiation forbids a single strategy from being dominant against all others in different negotiation scenarios. To overcome this, one approach is to use mixture of experts, but at the same time one problem of this method is the selection of experts, as this approach is limited by the competency of the experts selected. Another problem with most negotiation strategies is their incapability of adapting to dynamic variation of the opponent’s behaviour within a single negotiation session resulting in poor performance. This work focuses on both, solving the problem of expert selection and adapting to the opponent’s behaviour with our Autonomous Negotiating Agent Framework. This framework allows real-time classification of opponent’s behaviour and provides a mechanism to select, switch or combine strategies within a single negotiation session. Additionally, our framework has a reviewer component which enab...
Data Mining for Social Robotics: Toward Autonomously Social Robots
This book explores an approach to social robotics based solely on autonomous unsupervised techniq... more This book explores an approach to social robotics based solely on autonomous unsupervised techniques and positions it within a structured exposition of related research in psychology, neuroscience, HRI, and data mining. The authors present an autonomous and developmental approach that allows the robot to learn interactive behavior by imitating humans using algorithms from time-series analysis and machine learning. The first part provides a comprehensive and structured introduction to time-series analysis, change point discovery, motif discovery and causality analysis focusing on possible applicability to HRI problems. Detailed explanations of all the algorithms involved are provided with open-source implementations in MATLAB enabling the reader to experiment with them. Imitation and simulation are the key technologies used to attain social behavior autonomously in the proposed approach. Part two gives the reader a wide overview of research in these areas in psychology, and ethology....
MC^2 : An Integrated Toolbox for Change, Causality and Motif Discovery
Time series are being generated continuously from all kinds of human endeavors. The ubiquity of t... more Time series are being generated continuously from all kinds of human endeavors. The ubiquity of time-series data generates a need for data mining and pattern discovery algorithms targeting this data format which is becoming of ever increasing importance. Three basic problems in mining time-series data are change point discovery, causality discovery and motif discovery. This paper presents an integrated toolbox that can be used to perform any of these tasks on multidimensional real-valued time-series using state of the art algorithms. The proposed toolbox provides practitioners in time-series analysis and data mining with several tools useful for data generation, preprocessing, modeling evaluation and mining of long sequences. The paper also reports real world applications that uses the toolbox in HRI, physiological signal processing, and human behavior modeling and understanding.
Multidimensional Permutation Entropy for Constrained Motif Discovery
Constrained motif discovery was proposed as an unsupervised method for efficiently discovering in... more Constrained motif discovery was proposed as an unsupervised method for efficiently discovering interesting recurrent patterns in time-series. The de-facto standard way to calculate the required constraint on motif occurrence locations is change point discovery. This paper proposes the use of time-series complexity for finding the constraint and shows that the proposed approach can achieve higher accuracy in localizing motif occurrences and approximately the same accuracy for discovering different motifs at three times the speed of change point discovery. Moreover, the paper proposes a new extension of the permutation entropy for estimating time-series complexity to multi-dimensional time-series and shows that the proposed extension outperforms the state-of-the-art multi-dimensional permutation entropy approach both in speed and usability as a motif discovery constraint.
Spoken language identification is the process by which the language in a spoken utterance is reco... more Spoken language identification is the process by which the language in a spoken utterance is recognized automatically. Spoken language identification is commonly used in speech translation systems, in multi-lingual speech recognition, and in speaker diarization. In the current paper, spoken language identification based on deep learning (DL) and the i-vector paradigm is presented. Specifically, a comparative study is reported, consisting of experiments on language identification using deep neural networks (DNN) and convolutional neural networks (CNN). Also, the integration of the two methods into a complete system is investigated. Previous studies demonstrated the effectiveness of using DNN in spoken language identification. However, to date, the integration of CNN and i-vectors in language identification has not been investigated. The main advantage of using CNN is that fewer parameters are required compared to DNN. As a result, CNN is cheaper in terms of memory and the computation...
Previous research in HRI have shown that human's subjective evaluation of robot's abiliti... more Previous research in HRI have shown that human's subjective evaluation of robot's abilities affect the way people interact with robots. Given that one of the major challenges in learning from demonstration in robotics is the limited number of training examples that the demonstrator is usually willing to provide, it would be beneficial to design the interaction context in such a way to increase human's subjective evaluation of the robot's imitative skills. We propose back imitation as a way to achieve that goal. This paper reports the results of a preliminary study that was conducted to evaluate the effect of back imitation on human's subjective evaluation of the robot along several dimensions including imitation skill, motion human likeness, interaction quality, humanness and likability.
Conversational Informatics: A Data-Intensive Approach with Emphasis on Nonverbal Communication
This book covers an approach to conversational informatics which encompasses science and technolo... more This book covers an approach to conversational informatics which encompasses science and technology for understanding and augmenting conversation in the network age. A major challenge in engineering is to develop a technology for conveying not just messages but also underlying wisdom. Relevant theories and practices in cognitive linguistics and communication science, as well as techniques developed in computational linguistics and artificial intelligence, are discussed.
Spoken Language Identification Based on I-vectors and Conditional Random Fields
The task of an automatic language identification (LID) system is to automatically identify the la... more The task of an automatic language identification (LID) system is to automatically identify the language in a spoken utterance. Language identification can be applied as front-end to speech-to-speech translation systems, in speaker diarization, and at call centers to automatically route incoming calls to appropriate native speaker operators. In the current study, a method for automatic language identification based on i-vector paradigm and conditional random fields (CRF) is presented. CRF belong to discriminative classifiers and use an exponential distribution to model a sequence given the observation sequence. This allows non-independent observations, and allows also non-local dependencies between state and observation. When the proposed method is evaluated on the NIST 2015 i-vector Machine Learning Challenge task for the recognition of 50 in-set languages, a 3.7% equal error rate (EER) (i.e., miss probability equal to false alarms) is achieved. Using support vector machines (SVM) a...
NegMAS: A Platform for Situated Negotiations
Recent Advances in Agent-based Negotiation
The Proceedings of the 2nd International Conference on Industrial Application Engineering 2015
Learning from Demonstration is an important technology for the new wave of robots that are envisi... more Learning from Demonstration is an important technology for the new wave of robots that are envisioned to work side-by-side with workers in factories as well as social robots. Most available techniques for learning from demonstration rely on the existence of a training set of demonstrations that is assumed to be pre-segmented and is usually processed in batch. Another problem with most available methods is the need to set the model complexity used to model the motion. In this paper, we propose a solution to both problems based on incremental piecewise linear segmentation of the motion using an extension of the SWAB algorithm. Evaluation experiments show that the proposed method is able to generate motion models with adequate complexity without the need for model comparison methods and assuming incremental streaming of demonstrations rather than the availability of the complete training set in advance. The proposed method is applicable not only to robot learning from demonstration but to other industrial applications in which an accurate model of time variant data needs to be built from streaming input.
Concurrent local negotiations with a global utility function: a greedy approach
Autonomous Agents and Multi-Agent Systems
FastVOI: Efficient Utility Elicitation During Negotiations
Lecture Notes in Computer Science
Autonomous Negotiation is a promising technology that allows individuals and institutions to redu... more Autonomous Negotiation is a promising technology that allows individuals and institutions to reduce the burden and cost of negotiating win-win agreements. A common challenge in practical applications is the inability or high cost of finding the utility value for each possible outcome of the negotiation before it even starts. Earlier work on utility elicitation during negotiations tried to avoid the need of full revelation of the utility function to the agent by interleaving elicitation and negotiation actions. This paper proposes an efficient elicitation algorithm that allows the agent to achieve similar utility at orders of magnitude higher speed compared with the state-of-the-art algorithm.
Optimal Deterministic Time-Based Policy in Automated Negotiation
PRIMA 2020: Principles and Practice of Multi-Agent Systems
Supply Chain Management World
PRIMA 2019: Principles and Practice of Multi-Agent Systems
NegMAS: A Platform for Automated Negotiations
PRIMA 2020: Principles and Practice of Multi-Agent Systems
Primitive activity recognition from short sequences of sensory data
Applied Intelligence
Activity recognition from mobile device sensors and wearables is attracting more attention from t... more Activity recognition from mobile device sensors and wearables is attracting more attention from the research community due to the widespread adoption of these devices and the unique opportunity they provide for understanding user’s behavior leading to novel services and improvements in the delivery of existing ones. Approaches to tackle this problem either rely on predefined statistical features of sensor data streams or feature learning with the latter providing higher accuracies in most cases. Deep learning methods proved more effective than traditional approaches to feature learning in multiple studies. This paper presents a novel end-to-end trainable deep architecture that utilizes multiple convolutional neural networks (CNN), late fusion and extensive layer bypassing. The proposed method can easily accommodate multiple sensors and signal representations. The proposed approach is validated on eight publicly available datasets using a variety of evaluation conditions showing that it outperforms state-of-the-art methods in six of them.
Why Should We Imitate Robots? Effect of Back Imitation on Judgment of Imitative Skill
International Journal of Social Robotics
IEICE Transactions on Information and Systems
Activity recognition from sensors is a classification problem over time-series data. Some researc... more Activity recognition from sensors is a classification problem over time-series data. Some research in the area utilize time and frequency domain handcrafted features that differ between datasets. Another categorically different approach is to use deep learning methods for feature learning. This paper explores a middle ground in which an off-theshelf feature extractor is used to generate a large number of candidate timedomain features followed by a feature selector that was designed to reduce the bias toward specific classification techniques. Moreover, this paper advocates the use of features that are mostly insensitive to sensor orientation and show their applicability to the activity recognition problem. The proposed approach is evaluated using six different publicly available datasets collected under various conditions using different experimental protocols and shows comparable or higher accuracy than state-of-the-art methods on most datasets but usually using an order of magnitude fewer features.
Learning from Demonstration
Advanced Information and Knowledge Processing, 2015
ABSTRACT
Uploads
Papers by Yasser Mohammad