Showing 1–10 of 10 results for author: Chen, R L

A Blueprint Architecture of Compound AI Systems for Enterprise

Authors: Eser Kandogan, Sajjadur Rahman, Nikita Bhutani, Dan Zhang, Rafael Li Chen, Kushan Mitra, Sairam Gurajada, Pouya Pezeshkpour, Hayate Iso, Yanlin Feng, Hannah Kim, Chen Shen, Jin Wang, Estevam Hruschka

Abstract: Large Language Models (LLMs) have showcased remarkable capabilities surpassing conventional NLP challenges, creating opportunities for use in production use cases. Towards this goal, there is a notable shift to building compound AI systems, wherein LLMs are integrated into an expansive software infrastructure with many components like models, retrievers, databases and tools. In this paper, we intr… ▽ More Large Language Models (LLMs) have showcased remarkable capabilities surpassing conventional NLP challenges, creating opportunities for use in production use cases. Towards this goal, there is a notable shift to building compound AI systems, wherein LLMs are integrated into an expansive software infrastructure with many components like models, retrievers, databases and tools. In this paper, we introduce a blueprint architecture for compound AI systems to operate in enterprise settings cost-effectively and feasibly. Our proposed architecture aims for seamless integration with existing compute and data infrastructure, with ``stream'' serving as the key orchestration concept to coordinate data and instructions among agents and other components. Task and data planners, respectively, break down, map, and optimize tasks and data to available agents and data sources defined in respective registries, given production constraints such as accuracy and latency. △ Less

Submitted 1 June, 2024; originally announced June 2024.

Comments: Compound AI Systems Workshop at the Data+AI Summit 2024

MEGAnno+: A Human-LLM Collaborative Annotation System

Authors: Hannah Kim, Kushan Mitra, Rafael Li Chen, Sajjadur Rahman, Dan Zhang

Abstract: Large language models (LLMs) can label data faster and cheaper than humans for various NLP tasks. Despite their prowess, LLMs may fall short in understanding of complex, sociocultural, or domain-specific context, potentially leading to incorrect annotations. Therefore, we advocate a collaborative approach where humans and LLMs work together to produce reliable and high-quality labels. We present M… ▽ More Large language models (LLMs) can label data faster and cheaper than humans for various NLP tasks. Despite their prowess, LLMs may fall short in understanding of complex, sociocultural, or domain-specific context, potentially leading to incorrect annotations. Therefore, we advocate a collaborative approach where humans and LLMs work together to produce reliable and high-quality labels. We present MEGAnno+, a human-LLM collaborative annotation system that offers effective LLM agent and annotation management, convenient and robust LLM annotation, and exploratory verification of LLM labels by humans. △ Less

Submitted 27 February, 2024; originally announced February 2024.

Comments: EACL 2024 Demo

MEGAnno: Exploratory Labeling for NLP in Computational Notebooks

Authors: Dan Zhang, Hannah Kim, Rafael Li Chen, Eser Kandogan, Estevam Hruschka

Abstract: We present MEGAnno, a novel exploratory annotation framework designed for NLP researchers and practitioners. Unlike existing labeling tools that focus on data labeling only, our framework aims to support a broader, iterative ML workflow including data exploration and model development. With MEGAnno's API, users can programmatically explore the data through sophisticated search and automated sugges… ▽ More We present MEGAnno, a novel exploratory annotation framework designed for NLP researchers and practitioners. Unlike existing labeling tools that focus on data labeling only, our framework aims to support a broader, iterative ML workflow including data exploration and model development. With MEGAnno's API, users can programmatically explore the data through sophisticated search and automated suggestion functions and incrementally update task schema as their project evolve. Combined with our widget, the users can interactively sort, filter, and assign labels to multiple items simultaneously in the same notebook where the rest of the NLP project resides. We demonstrate MEGAnno's flexible, exploratory, efficient, and seamless labeling experience through a sentiment analysis use case. △ Less

Submitted 8 January, 2023; originally announced January 2023.

Comments: Data Science with Human-in-the-loop (DaSH) @ EMNLP 2022. Demo: https://meganno.github.io

Data-Driven Robust Optimization Using Scenario-Induced Uncertainty Sets

Authors: Meysam Cheramin, Richard Li-Yang Chen, Jianqiang Cheng, Ali Pinar

Abstract: Uncertainty sets are at the heart of robust optimization (RO) because they play a key role in determining the RO models' tractability, robustness, and conservativeness. Different types of uncertainty sets have been proposed that model uncertainty from various perspectives. Among them, polyhedral uncertainty sets are widely used due to their simplicity and flexible structure to model the underlying… ▽ More Uncertainty sets are at the heart of robust optimization (RO) because they play a key role in determining the RO models' tractability, robustness, and conservativeness. Different types of uncertainty sets have been proposed that model uncertainty from various perspectives. Among them, polyhedral uncertainty sets are widely used due to their simplicity and flexible structure to model the underlying uncertainty. However, the conventional polyhedral uncertainty sets present certain disadvantages; some are too conservative while others lead to computationally expensive RO models. This paper proposes a systematic approach to develop data-driven polyhedral uncertainty sets that mitigate these drawbacks. The proposed uncertainty sets are polytopes induced by a given set of scenarios, capture correlation information between uncertain parameters, and allow for direct trade-offs between tractability and conservativeness issue of conventional polyhedral uncertainty sets. To develop these uncertainty sets, we use principal component analysis (PCA) to transform the correlated scenarios into their uncorrelated principal components and to shrink the uncertainty space dimensionality. Thus, decision-makers can use the number of the leading principal components as a tool to trade-off tractability, conservativeness, and robustness of RO models. We quantify the quality of the lower bound of a static RO problem with a scenario-induced uncertainty set by deriving a theoretical bound on the optimality gap. Additionally, we derive probabilistic guarantees for the performance of the proposed scenario-induced uncertainty sets by developing explicit lower bounds on the number of scenarios. Finally, we demonstrate the practical applicability of the proposed uncertainty sets to trade-off tractability, robustness, and conservativeness by examining a range of knapsack and power grid problems. △ Less

Submitted 11 July, 2021; originally announced July 2021.

Contingency-Constrained Unit Commitment With Intervening Time for System Adjustments

Authors: Zhaomiao Guo, Richard Li-Yang Chen, Neng Fan, Jean-Paul Watson

Abstract: The N-1-1 contingency criterion considers the con- secutive loss of two components in a power system, with intervening time for system adjustments. In this paper, we consider the problem of optimizing generation unit commitment (UC) while ensuring N-1-1 security. Due to the coupling of time periods associated with consecutive component losses, the resulting problem is a very large-scale mixed-inte… ▽ More The N-1-1 contingency criterion considers the con- secutive loss of two components in a power system, with intervening time for system adjustments. In this paper, we consider the problem of optimizing generation unit commitment (UC) while ensuring N-1-1 security. Due to the coupling of time periods associated with consecutive component losses, the resulting problem is a very large-scale mixed-integer linear optimization model. For efficient solution, we introduce a novel branch-and-cut algorithm using a temporally decomposed bilevel separation oracle. The model and algorithm are assessed using multiple IEEE test systems, and a comprehensive analysis is performed to compare system performances across different contingency criteria. Computational results demonstrate the value of considering intervening time for system adjustments in terms of total cost and system robustness. △ Less

Submitted 18 April, 2016; originally announced April 2016.

Comments: 8 pages, 5 figures

Efficient Representation of Uncertainty for Stochastic Economic Dispatch

Authors: Cosmin Safta, Richard L. -Y. Chen, Habib N. Najm, Ali Pinar, Jean-Paul Watson

Abstract: Stochastic economic dispatch models address uncertainties in forecasts of renewable generation output by considering a finite number of realizations drawn from a stochastic process model, typically via Monte Carlo sampling. Accurate evaluations of expectations or higher-order moments for quantities of interest, e.g., generating cost, can require a prohibitively large number of samples. We propose… ▽ More Stochastic economic dispatch models address uncertainties in forecasts of renewable generation output by considering a finite number of realizations drawn from a stochastic process model, typically via Monte Carlo sampling. Accurate evaluations of expectations or higher-order moments for quantities of interest, e.g., generating cost, can require a prohibitively large number of samples. We propose an alternative to Monte Carlo sampling based on Polynomial Chaos expansions. These representations are based on sparse quadrature methods, and enable accurate propagation of uncertainties in model parameters. We also investigate a method based on Karhunen-Loeve expansions that enables us to efficiently represent uncertainties in renewable energy generation. Considering expected production cost, we demonstrate that the proposed approach can yield several orders of magnitude reduction in computational cost for solving stochastic economic dispatch relative to Monte Carlo sampling, for a given target error threshold. △ Less

Submitted 21 August, 2015; originally announced August 2015.

Comments: arXiv admin note: text overlap with arXiv:1407.2232

Toward Using Surrogates to Accelerate Solution of Stochastic Electricity Grid Operations Problems

Authors: Cosmin Safta, Richard L. Chen, Habib N. Najm, Ali Pinar, Jean-paul watson

Abstract: Stochastic unit commitment models typically handle uncertainties in forecast demand by considering a finite number of realizations from a stochastic process model for loads. Accurate evaluations of expectations or higher moments for the quantities of interest require a prohibitively large number of model evaluations. In this paper we propose an alternative approach based on using surrogate models… ▽ More Stochastic unit commitment models typically handle uncertainties in forecast demand by considering a finite number of realizations from a stochastic process model for loads. Accurate evaluations of expectations or higher moments for the quantities of interest require a prohibitively large number of model evaluations. In this paper we propose an alternative approach based on using surrogate models valid over the range of the forecast uncertainty. We consider surrogate models based on Polynomial Chaos expansions, constructed using sparse quadrature methods. Considering expected generation cost, we demonstrate the approach can lead to several orders of magnitude reduction in computational cost relative to using Monte Carlo sampling on the original model, for a given target error threshold. △ Less

Submitted 1 June, 2014; originally announced July 2014.

Contingency-Constrained Unit Commitment with Post-Contingency Corrective Recourse

Authors: Richard Li-Yang Chen, Neng Fan, Ali Pinar, Jean-Paul Watson

Abstract: We consider the problem of minimizing costs in the generation unit commitment problem, a cornerstone in electric power system operations, while enforcing an N-k-e reliability criterion. This reliability criterion is a generalization of the well-known $N$-$k$ criterion, and dictates that at least $(1-e_ j)$ fraction of the total system demand must be met following the failures of $k$ or fewer syste… ▽ More We consider the problem of minimizing costs in the generation unit commitment problem, a cornerstone in electric power system operations, while enforcing an N-k-e reliability criterion. This reliability criterion is a generalization of the well-known $N$-$k$ criterion, and dictates that at least $(1-e_ j)$ fraction of the total system demand must be met following the failures of $k$ or fewer system components. We refer to this problem as the Contingency-Constrained Unit Commitment problem, or CCUC. We present a mixed-integer programming formulation of the CCUC that accounts for both transmission and generation element failures. We propose novel cutting plane algorithms that avoid the need to explicitly consider an exponential number of contingencies. Computational studies are performed on several IEEE test systems and a simplified model of the Western US interconnection network, which demonstrate the effectiveness of our proposed methods relative to current state-of-the-art. △ Less

Submitted 10 April, 2014; originally announced April 2014.

Contingency-Risk Informed Power System Design

Authors: Richard Li-Yang Chen, Amy Cohn, Neng Fan, Ali Pinar

Abstract: We consider the problem of designing (or augmenting) an electric power system at a minimum cost such that it satisfies the N-k-e survivability criterion. This survivability criterion is a generalization of the well-known N-k criterion, and it requires that at least (1- e_j) fraction of the total demand to be met after failures of up to j components, for j=1,...,k. The network design problem adds a… ▽ More We consider the problem of designing (or augmenting) an electric power system at a minimum cost such that it satisfies the N-k-e survivability criterion. This survivability criterion is a generalization of the well-known N-k criterion, and it requires that at least (1- e_j) fraction of the total demand to be met after failures of up to j components, for j=1,...,k. The network design problem adds another level of complexity to the notoriously hard contingency analysis problem, since the contingency analysis is only one of the requirements for the design optimization problem. We present a mixed-integer programming formulation of this problem that takes into account both transmission and generation expansion. We propose an algorithm that can avoid combinatorial explosion in the number of contingencies, by seeking vulnerabilities in intermediary solutions and constraining the design space accordingly. Our approach is built on our ability to identify such system vulnerabilities quickly. Our empirical studies on modified instances from the IEEE 30-bus and IEEE 57-bus systems show the effectiveness of our methods. We were able to solve the transmission and generation expansion problems for k=4 under 2 minutes, while other approaches failed to provide a solution at the end of 2 hours. △ Less

Submitted 3 May, 2013; originally announced May 2013.

Comments: arXiv admin note: substantial text overlap with arXiv:1201.1530

N-k-e Survivable Power System Design

Authors: Richard Li-Yang Chen, Amy Cohn, Neng Fan, Ali Pinar

Abstract: We consider the problem of designing (or augmenting) an electric power system such that it satisfies the N-k-e survivability criterion while minimizing total cost. The survivability criterion requires that at least (1-e) fraction of the total demand can still be met even if any k (or fewer) of the system components fail. We formulate this problem, taking into account both transmission and generati… ▽ More We consider the problem of designing (or augmenting) an electric power system such that it satisfies the N-k-e survivability criterion while minimizing total cost. The survivability criterion requires that at least (1-e) fraction of the total demand can still be met even if any k (or fewer) of the system components fail. We formulate this problem, taking into account both transmission and generation expansion planning, as a mixed-integer program. Two algorithms are designed and tested on modified instances from the IEEE-30-Bus and IEEE- 57-Bus systems. △ Less

Submitted 6 January, 2012; originally announced January 2012.