About

Alex Doboli is a Professor in the Department of Electrical and Computer Engineering at Stony Brook University. His research focuses on computer-aided design (CAD) for electronic circuits and systems, as well as the specification, modeling, and synthesis of analog and mixed-signal circuits. Professor Doboli's work contributes to advancing methodologies and tools that support the design and development of complex electronic systems, emphasizing the integration of analog and digital components. He is based in the Light Engineering building at Stony Brook University and can be contacted via email at Alex.Doboli@stonybrook.edu.

Research topics

• Computer Science
• Artificial Intelligence
• Psychology
• Data science
• Natural Language Processing
• Machine Learning
• World Wide Web
• Management science
• Operations research
• Marketing
• Mathematics
• Social psychology
• Economics
• Engineering
• Mathematics education
• Knowledge management
• Risk analysis (engineering)
• Business
• Epistemology

Selected publications

• A Stacked Multi-Layered Perceptron - LLM Model for Extracting the Relations in Textual Descriptions

  2025-03-17 · 4 citations

  article

  The automated extraction of the relations among the concepts in textual descriptions is important for problems that require creating implementations of those descriptions, e.g., synthesizing engineering designs and computer code. However, relation extraction remains challenging despite significant recent progress in Natural Language Processing. This paper proposes a novel, two-layered method to create structural representations of the relations in a text. Inspired by work in neuroscience, an upper layer implemented as a multi-layer perceptron models the behavior of closed class words (words with a fixed meaning), like prepositions and conjunctions. The lower layer prompts a Large Language Model to extract the nouns and verbs, which are then introduced into the structural representation produced by the upper layer. Experiments show an average improvement of 13.6% as compared to using only LLMs.

  PublisherDOI

• Using a Symbolic Knowledge Graph to Address LLM Limitations in Analog Circuit Topology Generation

  2025-01-06 · 4 citations

  articleSenior author

  Analog circuit topology synthesis has been shown to be hard arguably due to the difficulty to formulate it either as a regression or as an optimization problem, which are currently the two main synthesis approaches. This paper proposes a novel analog circuit topology synthesis methodology that combines the capabilities of Deep Neural Models with the advantages of symbolic representations. It integrates Large Language Models (LLMs) with Symbolic Knowledge Graphs (SKGs): an LLM offers broad descriptions of the topological modules that can be incorporated into the final schematics, and an SKG incorporates all design details needed to complete the topological modules and other related circuitry of a circuit schematics. The SKG also supports the causal reasoning flow through which the modules indicated by the LLM are selected and integrated with each other.

  PublisherDOI

• An Overview of Large Language Models and a Novel, Large Language Model-Based Cognitive Architecture for Solving Open-Ended Problems

  Machine Learning and Knowledge Extraction · 2025-11-01

  articleOpen accessSenior authorCorresponding

  Large Language Models (LLMs) offer new opportunities to devise automated implementation generation methods that can tackle problem solving beyond traditional methods, which usually require algorithmic specifications and use only static domain knowledge. LLMs can support devising new methods to support activities in tackling open-ended problems, like problem framing, exploring possible solving approaches, feature elaboration and combination, advanced implementation assessment, and handling unexpected situations. This paper presents a detailed overview of the current work on LLMs, including model prompting, retrieval-augmented generation (RAG), and reinforcement learning. It then proposes a novel, LLM-based Cognitive Architecture (CA) to generate programming code starting from verbal discussions in natural language, a particular kind of problem-solving activity. The CA uses four strategies, three top-down and one bottom-up, to elaborate, adaptively process, memorize, and learn. Experiments are devised to study the CA performance, e.g., convergence rate, semantic fidelity, and code correctness.

  PublisherDOI

• Optimized Interdisciplinary Research Team Formation Using a Genetic Algorithm and Publication Metadata Records

  AI · 2025-07-30

  articleOpen accessSenior author

  Forming interdisciplinary research teams is challenging, especially when the pool of candidates is large and/or the addressed research projects require multi-disciplinary expertise. Based on their previous research outputs, like published work, a data-driven team formation procedure selects the researchers that are likely to work well together while covering all areas and offering all skills required by the multi-disciplinary topic. The description of the research team formation problem proposed in this paper uses novel quantitative metrics about the team candidates computed from bibliographic metadata records. The proposed methodology first analyzes the metadata fields that provide useful information and then computes four synthetic indicators regarding candidates’ skills and their interpersonal traits. Interdisciplinary teams are formed by solving a complex combinatorial multi-objective weighted set cover optimization problem, defined as equations involving the synthetic indicators. Problem solving uses the NSGA-II genetic algorithm. The proposed methodology is validated and compared with other similar approaches using a dataset on researchers from Politehnica University of Timisoara extracted from the IEEE Xplore database. Experimental results show that the method can identify potential research teams in situations for which other related algorithms fail.

  PublisherOA PDFDOI

• Two Large Language Model-based Methods to Validate Open-Ended Problem Solving in Teams

  2025-11-25

  articleSenior author

  This paper describes two Large Language Model (LLM)-based methods to validate programming solutions devised in small teams for open-ended problems. The methods address the static mode, when validation is conducted at the end of problem solving, and a real-time mode, if validation is repeatedly performed at consecutive time intervals. The validation methods automatically prompt an LLM to find inconsistencies between the problem description, the user intentions verbally expressed through dialog, and the created programming code. Experiments studied the performance of the two validation methods.

  PublisherDOI

• Modeling and Adaptive Resource Management for Voice-Based Speaker and Emotion Identification Through Smart Badges

  Electronics · 2025-02-17

  articleOpen accessSenior authorCorresponding

  The number of new applications addressing human activities in social settings, like groups and organizations, is on the rise. Devising an effective data collection infrastructure is critical for such applications. This paper describes a computational model and the related algorithms to design a sociometric badge for efficient data collection in applications in which speaker and emotion recognition and tracking are essential. A new computational model describes the characteristics of verbal and emotional interactions in a group. To address the requirements of changing group interactions, a self-adaptation module optimizes badge resource management to minimize data loss and modeling errors. Experiments considered scenarios for slow and regular shifts in group interactions. The proposed self-adaptation method reduces data loss by 51% to 90%, modeling errors by 28% to 44%, and computing load by 38% to 52%.

  PublisherOA PDFDOI

• Towards Semantic Classification: An Experimental Study on Automated Understanding of the Meaning of Verbal Utterances

  2025-01-06

  article

  Computationally understanding the meaning of verbal discussions in groups can improve group effectiveness by optimizing their interactions and allocating the needed physical and Cyber resources. Still, it is unknown to what degree the existing Machine Learning (ML) methods can automatically detect the type of verbal utterances, as a preliminary step towards automated meaning understanding. This paper presents a comprehensive experimental study of the performance of the main ML methods in classifying verbal utterances depending on their role during solving programming exercises. A model for interpretable classification using decision trees is also offered. The paper summarizes a set of requirements that new semantic classifiers must satisfy, as current ML methods are likely insufficient for the task. These requirements were experimentally validated.

  PublisherDOI

• Using Knowledge Graph Dynamics to Describe Group Ideation Effectiveness in Problem Solving

  2025-06-30

  article

  Groups engaged in brainstorming undergo several types of ideation processes, some more effective than others. Current methods for characterizing the knowledge expressed in a group and its dynamics include embedding models and cosine distance between idea embeddings that do not find explicit semantic connections, or knowledge graphs that represent semantic connections but need large amounts of data to extract. We are proposing a novel idea graph model that combines embedding models and a graph representation of knowledge, together with a novel group problem-solving model. Using the graph representation and the evaluation measures, we identified and characterized more effective groups from experimental data and verified the main hypotheses derived from the theoretical model.

  PublisherDOI

• An Experimental Study on the Interpretability of Transformer Models for Dialog Understanding

  2025-05-05

  articleSenior author

  Identifying the intention of an utterance (e.g., spoken sentence) is part of semantic understanding. Transformer models offer promising performance for intention identification, however model interpretability remains low. This paper presents a comprehensive experimental study on the interpretability of BERT-like models, such as DistilBERT, for spoken dialog intention understanding. A detailed discussion explains the main features used in intention classification. This insight can be a starting point to devise more interpretable transformer models.

  PublisherDOI

• diaLogic: A Multi-Modal Framework for Automated Team Behavior Modeling Based on Speech Acquisition

  Multimodal Technologies and Interaction · 2025-03-10

  articleOpen accessSenior authorCorresponding

  This paper presents diaLogic, a humans-in-the-loop system for modeling the behavior of teams during collective problem solving. Team behavior is modeled using multi-modal data about cognition, social interactions, and emotions acquired from speech inputs. The system includes methods for speaker diarization, speaker interaction characterization, speaker emotion recognition, and speech-to-text conversion. Hypotheses about the invariant and differentiated aspects of teams are extracted using the similarities and dissimilarities of their behavior over time. Hypothesis extraction, a novel contribution of this work, uses a method to identify the clauses and concepts in each spoken sentence. Experiments present system performance for a broad set of cases of team behavior during problem solving. The average errors of the various methods are between 6% and 21%. The system can be used in a broad range of applications, from education to team research and therapy.

  PublisherOA PDFDOI

Recent grants

• Major Collaborative Proposal: Understanding the Creative Design Process: A Novel Cognitive Model based on Behavioral Experiments in Circuit Design and Architecture

  NSF · $257k · 2009–2014

Frequent coauthors

• Ranga Vemuri

  University of Cincinnati

  50 shared

• Cristian Ferent

  Stony Brook University

  48 shared

• Varun Subramanian

  Arizona State University

  47 shared

• Simona Doboli

  Polytechnic University of Timişoara

  45 shared

• Michael Gilberti

  35 shared

• Fanshu Jiao

  Stony Brook University

  34 shared

• Anurag Umbarkar

  Stony Brook University

  34 shared

• Hua Tang

  University of Minnesota, Duluth

  29 shared

Education

• PhD, Electrical and Computer Engineering and Computer Science

  University of Cincinnati

  2000

Awards & honors

• IBM Partnership Award (2001)
• "Traian Lalescu" Award, "Politehnica" University Timisoara (…