Showing 1–11 of 11 results for author: Papadimitriou, D

Bayesian Constraint Inference from User Demonstrations Based on Margin-Respecting Preference Models

Authors: Dimitris Papadimitriou, Daniel S. Brown

Abstract: It is crucial for robots to be aware of the presence of constraints in order to acquire safe policies. However, explicitly specifying all constraints in an environment can be a challenging task. State-of-the-art constraint inference algorithms learn constraints from demonstrations, but tend to be computationally expensive and prone to instability issues. In this paper, we propose a novel Bayesian… ▽ More It is crucial for robots to be aware of the presence of constraints in order to acquire safe policies. However, explicitly specifying all constraints in an environment can be a challenging task. State-of-the-art constraint inference algorithms learn constraints from demonstrations, but tend to be computationally expensive and prone to instability issues. In this paper, we propose a novel Bayesian method that infers constraints based on preferences over demonstrations. The main advantages of our proposed approach are that it 1) infers constraints without calculating a new policy at each iteration, 2) uses a simple and more realistic ranking of groups of demonstrations, without requiring pairwise comparisons over all demonstrations, and 3) adapts to cases where there are varying levels of constraint violation. Our empirical results demonstrate that our proposed Bayesian approach infers constraints of varying severity, more accurately than state-of-the-art constraint inference methods. △ Less

Submitted 4 March, 2024; originally announced March 2024.

A primal-dual backward reflected forward splitting algorithm for structured monotone inclusions

Authors: Vu Cong Bang, Dimitri Papadimitriou, Vu Xuan Nham

Abstract: We propose a primal-dual backward reflected forward splitting method for solving structured primal-dual monotone inclusion in real Hilbert space. The algorithm allows to use the inexact computations of the Lipschitzian and cocoercive operators. The strong convergence of the generated iterative sequence is proved under the strong monotonicity condition, whilst the weak convergence is formally prove… ▽ More We propose a primal-dual backward reflected forward splitting method for solving structured primal-dual monotone inclusion in real Hilbert space. The algorithm allows to use the inexact computations of the Lipschitzian and cocoercive operators. The strong convergence of the generated iterative sequence is proved under the strong monotonicity condition, whilst the weak convergence is formally proved under several conditioned used in the literature. An application to a structured minimization problem is supported. △ Less

Submitted 10 January, 2024; originally announced January 2024.

Comments: accepted for publication in AMV

Constraint Inference in Control Tasks from Expert Demonstrations via Inverse Optimization

Authors: Dimitris Papadimitriou, **gqi Li

Abstract: Inferring unknown constraints is a challenging and crucial problem in many robotics applications. When only expert demonstrations are available, it becomes essential to infer the unknown domain constraints to deploy additional agents effectively. In this work, we propose an approach to infer affine constraints in control tasks after observing expert demonstrations. We formulate the constraint infe… ▽ More Inferring unknown constraints is a challenging and crucial problem in many robotics applications. When only expert demonstrations are available, it becomes essential to infer the unknown domain constraints to deploy additional agents effectively. In this work, we propose an approach to infer affine constraints in control tasks after observing expert demonstrations. We formulate the constraint inference problem as an inverse optimization problem, and we propose an alternating optimization scheme that infers the unknown constraints by minimizing a KKT residual objective. We demonstrate the effectiveness of our method in a number of simulations, and show that our method can infer less conservative constraints than a recent baseline method, while maintaining comparable safety guarantees. △ Less

Submitted 4 December, 2023; v1 submitted 6 April, 2023; originally announced April 2023.

An inertial ADMM for a class of nonconvex composite optimization with nonlinear coupling constraints

Authors: Le Thi Khanh Hien, Dimitri Papadimitriou

Abstract: In this paper, we propose an inertial alternating direction method of multipliers for solving a class of non-convex multi-block optimization problems with \emph{nonlinear coupling constraints}. Distinctive features of our proposed method, when compared with other alternating direction methods of multipliers for solving non-convex problems with nonlinear coupling constraints, include: (i) we apply… ▽ More In this paper, we propose an inertial alternating direction method of multipliers for solving a class of non-convex multi-block optimization problems with \emph{nonlinear coupling constraints}. Distinctive features of our proposed method, when compared with other alternating direction methods of multipliers for solving non-convex problems with nonlinear coupling constraints, include: (i) we apply the inertial technique to the update of primal variables and (ii) we apply a non-standard update rule for the multiplier by scaling the multiplier by a factor before moving along the ascent direction where a relaxation parameter is allowed. Subsequential convergence and global convergence are presented for the proposed algorithm. △ Less

Submitted 21 February, 2024; v1 submitted 21 December, 2022; originally announced December 2022.

Multiblock ADMM for nonsmooth nonconvex optimization with nonlinear coupling constraints

Authors: Le Thi Khanh Hien, Dimitri Papadimitriou

Abstract: This paper proposes a multiblock alternating direction method of multipliers for solving a class of multiblock nonsmooth nonconvex optimization problem with nonlinear coupling constraints. We employ a majorization minimization procedure in the update of each block of the primal variables. Subsequential and global convergence of the generated sequence to a critical point of the augmented Lagrangian… ▽ More This paper proposes a multiblock alternating direction method of multipliers for solving a class of multiblock nonsmooth nonconvex optimization problem with nonlinear coupling constraints. We employ a majorization minimization procedure in the update of each block of the primal variables. Subsequential and global convergence of the generated sequence to a critical point of the augmented Lagrangian are proved. We also establish iteration complexity and provide preliminary numerical results for the proposed algorithm. △ Less

Submitted 2 December, 2023; v1 submitted 19 January, 2022; originally announced January 2022.

Data-Driven Low-Rank Neural Network Compression

Authors: Dimitris Papadimitriou, Swayambhoo Jain

Abstract: Despite many modern applications of Deep Neural Networks (DNNs), the large number of parameters in the hidden layers makes them unattractive for deployment on devices with storage capacity constraints. In this paper we propose a Data-Driven Low-rank (DDLR) method to reduce the number of parameters of pretrained DNNs and expedite inference by imposing low-rank structure on the fully connected layer… ▽ More Despite many modern applications of Deep Neural Networks (DNNs), the large number of parameters in the hidden layers makes them unattractive for deployment on devices with storage capacity constraints. In this paper we propose a Data-Driven Low-rank (DDLR) method to reduce the number of parameters of pretrained DNNs and expedite inference by imposing low-rank structure on the fully connected layers, while controlling for the overall accuracy and without requiring any retraining. We pose the problem as finding the lowest rank approximation of each fully connected layer with given performance guarantees and relax it to a tractable convex optimization problem. We show that it is possible to significantly reduce the number of parameters in common DNN architectures with only a small reduction in classification accuracy. We compare DDLR with Net-Trim, which is another data-driven DNN compression technique based on sparsity and show that DDLR consistently produces more compressed neural networks while maintaining higher accuracy. △ Less

Submitted 12 July, 2021; originally announced July 2021.

Control of Unknown Nonlinear Systems with Linear Time-Varying MPC

Authors: Dimitris Papadimitriou, Ugo Rosolia, Francesco Borrelli

Abstract: We present a Model Predictive Control (MPC) strategy for unknown input-affine nonlinear dynamical systems. A non-parametric method is used to estimate the nonlinear dynamics from observed data. The estimated nonlinear dynamics are then linearized over time varying regions of the state space to construct an Affine Time Varying (ATV) model. Error bounds arising from the estimation and linearization… ▽ More We present a Model Predictive Control (MPC) strategy for unknown input-affine nonlinear dynamical systems. A non-parametric method is used to estimate the nonlinear dynamics from observed data. The estimated nonlinear dynamics are then linearized over time varying regions of the state space to construct an Affine Time Varying (ATV) model. Error bounds arising from the estimation and linearization procedure are computed by using sampling techniques. The ATV model and the uncertainty sets are used to design a robust Model Predictive Control (MPC) problem which guarantees safety for the unknown system with high probability. A simple nonlinear example demonstrates the effectiveness of the approach where commonly used linearization methods fail. △ Less

Submitted 8 October, 2020; v1 submitted 6 April, 2020; originally announced April 2020.

Comments: Fixed double citation in section 4, typos and combined two propositions into one

Molecular Modelling Combined with Advanced Chemistry for the Rational Design of Efficient Graphene Dispersing Agents

Authors: Konstantinia D. Papadimitriou, Emmanuel N. Skountzos, Sandra S. Gkermpoura, Ioannis Polyzos, Vlasis G. Mavrantzas, Costas Galiotisa, Constantinos Tsitsilianis

Abstract: Pyrene-functional PMMAs were prepared via ATRP-controlled polymerization and click reaction, as efficient dispersing agents for the exfoliation of few-layered graphene sheets (GS) in easily processable low boiling point chloroform. In parallel, detailed atomistic simulations showed fine dispersion of the GS/polymer hybrids in good agreement with the experiment. Moreover, the molecular dynamics sim… ▽ More Pyrene-functional PMMAs were prepared via ATRP-controlled polymerization and click reaction, as efficient dispersing agents for the exfoliation of few-layered graphene sheets (GS) in easily processable low boiling point chloroform. In parallel, detailed atomistic simulations showed fine dispersion of the GS/polymer hybrids in good agreement with the experiment. Moreover, the molecular dynamics simulations revealed interesting conformations (bridges, loops, dangling ends, free chains) of GS/polymer hybrids and allowed us to monitor their time evolution both in solution and in the polymer nanocomposite where the solvent molecules were replaced with PMMA chains. Microscopic information about these structures is very important for optimizing mechanical performance. It seems that the combination of atomistic simulation with advanced chemistry constitutes a powerful tool for the design of effective graphene dispersing agents that could be used for the production of graphene-based nanocomposites with tailor-made mechanical properties. △ Less

Submitted 3 May, 2018; originally announced May 2018.

Journal ref: ACS Macro Lett., 2016, 5, 24-29

Colloidal stabilization of graphene sheets by ionizable amphiphilic block copolymers in various media

Authors: Maria-Teodora Popescu, Dimitrios Tasis, Konstantinia D. Papadimitriou, Sandra Gkermpoura, Costas Galiotisac, Constantinos Tsitsilianis

Abstract: In this work, linear polystyrene-poly(2-vinylpyridine) (PS-b-P2VP) and heteroarm star PS_22P2VP_22 ionizable block copolymers were used as dispersing agents for the liquid-phase exfoliation of pristine graphene. Various strategies such as direct exfoliation, film hydration and phase transfer have been employed and compared. The best strategy involved a two-step process, namely, pre-exfoliation of… ▽ More In this work, linear polystyrene-poly(2-vinylpyridine) (PS-b-P2VP) and heteroarm star PS_22P2VP_22 ionizable block copolymers were used as dispersing agents for the liquid-phase exfoliation of pristine graphene. Various strategies such as direct exfoliation, film hydration and phase transfer have been employed and compared. The best strategy involved a two-step process, namely, pre-exfoliation of graphite in polymer/CHCl3 solutions followed by phase transfer to acidified water. High concentrations of stable aqueous suspensions of graphene flakes, highly enriched in monolayer structures, were then obtained by using the star-shaped copolymers as stabilizers. The as-prepared graphene/copolymer hybrids were used as a filler material in order to prepare functional polymer composites for mechanical reinforcement. Such copolymer-modified graphene sheets have proven to be efficient reinforcing agents of PVA, as a significant increase of storage modulus (145% higher than that of neat PVA) was achieved even at a low graphene weight fraction of 0.1 wt%. △ Less

Submitted 22 January, 2018; originally announced January 2018.

Journal ref: RSC Adv., 2015,5, 89447-89460

Feasibility study on distributed simulations of BGP

Authors: David Coudert, Luc Hogie, Aurélien Lancin, Dimitri Papadimitriou, Stéphane Pérennes, Issam Tahiri

Abstract: The Autonomous System (AS)-level topology of the Internet that currently comprises 40k ASs, is growing at a rate of about 10% per year. In these conditions, Border Gateway Protocol (BGP), the inter-domain routing protocol of the Internet starts to show its limits, among others in terms of the number of routing table entries it can dynamically process and control. To overcome this challenging situa… ▽ More The Autonomous System (AS)-level topology of the Internet that currently comprises 40k ASs, is growing at a rate of about 10% per year. In these conditions, Border Gateway Protocol (BGP), the inter-domain routing protocol of the Internet starts to show its limits, among others in terms of the number of routing table entries it can dynamically process and control. To overcome this challenging situation, the design but also the evaluation of alternative dynamic routing models and their comparison with BGP shall be performed by means of simulation. For this purpose, DRMSim, a Dynamic Routing Model Simulator, was developed that provides the means for large-scale simulations of various routing models including BGP. By means of this discrete-event simulator, execution of path-vector routing, e.g. BGP, and other compact routing models have been successfully performed on network topologies comprising more than ten thousand (abstract) nodes. However, to simulate dynamic routing schemes like BGP, DRMSim needs enhancements to support current Internet size (40k ASs) and even more by considering its evolution (up to 100k ASs). This paper proposes a feasibility study of the extension of DRMSim so as to support the Distributed Parallel Discrete Event paradigm. We first detail the possible distribution models and their associated communication overhead. Then, we analyze the communication overhead of such a distributed simulator by executing BGP on a partitioned topology according to different scenarios. Finally, we conclude on the feasibility of such a simulator by computing the expected additional time required by a distributed simulation of BGP compared to its sequential simulation. △ Less

Submitted 17 April, 2013; originally announced April 2013.

Report number: RR-8283

Feasibility study on distributed simulations of BGP

Authors: David Coudert, Luc Hogie, Aurélien Lancin, Dimitri Papadimitriou, Stéphane Pérennes, Issam Tahiri

Abstract: The Autonomous System (AS) topology of the Internet (up to 61k ASs) is growing at a rate of about 10% per year. The Border Gateway Protocol (BGP) starts to show its limits in terms of the number of routing table entries it can dynamically process and control. Due to the increasing routing information processing and storage, the same trend is observed for routing model simulators such as DRMSim spe… ▽ More The Autonomous System (AS) topology of the Internet (up to 61k ASs) is growing at a rate of about 10% per year. The Border Gateway Protocol (BGP) starts to show its limits in terms of the number of routing table entries it can dynamically process and control. Due to the increasing routing information processing and storage, the same trend is observed for routing model simulators such as DRMSim specialized in large-scale simulations of routing models. Therefore, DRMSim needs enhancements to support the current size of the Internet topology and its evolution (up to 100k ASs). To this end, this paper proposes a feasibility study of the extension of DRMSim so as to support the Distributed Parallel Discrete Event paradigm. We first detail the possible distribution models and their associated communication overhead. Then, we analyze this overhead by executing BGP on a partitioned topology according to different scenarios. Finally, we conclude on the feasibility of such a simulator by computing the expected additional time required by a distributed simulation of BGP compared to its sequential simulation. △ Less

Submitted 5 September, 2012; originally announced September 2012.

Comments: 26th ACM/IEEE/SCS Workshop on Principles of Advanced and Distributed Simulation (2012)