arXiv subjects

Mohammadreza Nazari

Publications and source records attributed to Mohammadreza Nazari.

8 recordsLinked to original sources

arXiv2022

Machine Learning to Generate Adjustable Dose Distributions in Head-and-Neck Cancer Radiation Therapy

In this work, we propose a Machine Learning model that generates an adjustable 3D dose distribution for external beam radiation therapy for head-and-neck cancer treatments. In contrast to existing Machine Learning methods that provide a single model, we create pairs of models for each organ-at-risk, namely lower-extreme and upper-extreme models. These model pairs for an organ-at-risk propose doses that give lower and higher doses to that organ-at-risk, while also encapsulating the dose trade-off to other organs-at-risk. By weighting and combining the model pairs for all organs-at-risk, we are able to dynamically create adjustable dose distributions that can be used, in real-time, to move doses between organs-at-risk, thereby customizing the dose distribution to the needs of a particular patient. We leverage a key observation that the training data set inherently contains the clinical trade-offs. We show that the adjustable distributions are able to provide reasonable clinical dose latitude in the trade-off of doses between organs-at-risk.

physics.med-ph↗

arXiv2020

AttendLight: Universal Attention-Based Reinforcement Learning Model for Traffic Signal Control

We propose AttendLight, an end-to-end Reinforcement Learning (RL) algorithm for the problem of traffic signal control. Previous approaches for this problem have the shortcoming that they require training for each new intersection with a different structure or traffic flow distribution. AttendLight solves this issue by training a single, universal model for intersections with any number of roads, lanes, phases (possible signals), and traffic flow. To this end, we propose a deep RL model which incorporates two attention models. The first attention model is introduced to handle different numbers of roads-lanes; and the second attention model is intended for enabling decision-making with any number of phases in an intersection. As a result, our proposed model works for any intersection configuration, as long as a similar configuration is represented in the training set. Experiments were conducted with both synthetic and real-world standard benchmark data-sets. The results we show cover intersections with three or four approaching roads; one-directional/bi-directional roads with one, two, and three lanes; different number of phases; and different traffic flows. We consider two regimes: (i) single-environment training, single-deployment, and (ii) multi-environment training, multi-deployment. AttendLight outperforms both classical and other RL-based approaches on all cases in both regimes.

cs.LG↗

arXiv2020

SONIA: A Symmetric Blockwise Truncated Optimization Algorithm

This work presents a new algorithm for empirical risk minimization. The algorithm bridges the gap between first- and second-order methods by computing a search direction that uses a second-order-type update in one subspace, coupled with a scaled steepest descent step in the orthogonal complement. To this end, partial curvature information is incorporated to help with ill-conditioning, while simultaneously allowing the algorithm to scale to the large problem dimensions often encountered in machine learning applications. Theoretical results are presented to confirm that the algorithm converges to a stationary point in both the strongly convex and nonconvex cases. A stochastic variant of the algorithm is also presented, along with corresponding theoretical guarantees. Numerical results confirm the strengths of the new approach on standard machine learning problems.

math.OC↗

arXiv2020

Scaling Up Quasi-Newton Algorithms: Communication Efficient Distributed SR1

In this paper, we present a scalable distributed implementation of the Sampled Limited-memory Symmetric Rank-1 (S-LSR1) algorithm. First, we show that a naive distributed implementation of S-LSR1 requires multiple rounds of expensive communications at every iteration and thus is inefficient. We then propose DS-LSR1, a communication-efficient variant that: (i) drastically reduces the amount of data communicated at every iteration, (ii) has favorable work-load balancing across nodes, and (iii) is matrix-free and inverse-free. The proposed method scales well in terms of both the dimension of the problem and the number of data points. Finally, we illustrate the empirical performance of DS-LSR1 on a standard neural network training task.

math.OC↗

arXiv2019

Multi-Agent Image Classification via Reinforcement Learning

We investigate a classification problem using multiple mobile agents capable of collecting (partial) pose-dependent observations of an unknown environment. The objective is to classify an image over a finite time horizon. We propose a network architecture on how agents should form a local belief, take local actions, and extract relevant features from their raw partial observations. Agents are allowed to exchange information with their neighboring agents to update their own beliefs. It is shown how reinforcement learning techniques can be utilized to achieve decentralized implementation of the classification problem by running a decentralized consensus protocol. Our experimental results on the MNIST handwritten digit dataset demonstrates the effectiveness of our proposed framework.

cs.LG↗

arXiv2019

Don't Forget Your Teacher: A Corrective Reinforcement Learning Framework

Although reinforcement learning (RL) can provide reliable solutions in many settings, practitioners are often wary of the discrepancies between the RL solution and their status quo procedures. Therefore, they may be reluctant to adapt to the novel way of executing tasks proposed by RL. On the other hand, many real-world problems require relatively small adjustments from the status quo policies to achieve improved performance. Therefore, we propose a student-teacher RL mechanism in which the RL (the "student") learns to maximize its reward, subject to a constraint that bounds the difference between the RL policy and the "teacher" policy. The teacher can be another RL policy (e.g., trained under a slightly different setting), the status quo policy, or any other exogenous policy. We formulate this problem using a stochastic optimization model and solve it using a primal-dual policy gradient algorithm. We prove that the policy is asymptotically optimal. However, a naive implementation suffers from high variance and convergence to a stochastic optimal policy. With a few practical adjustments to address these issues, our numerical experiments confirm the effectiveness of our proposed method in multiple GridWorld scenarios.

cs.LG↗

arXiv2018

Reinforcement Learning for Solving the Vehicle Routing Problem

We present an end-to-end framework for solving the Vehicle Routing Problem (VRP) using reinforcement learning. In this approach, we train a single model that finds near-optimal solutions for problem instances sampled from a given distribution, only by observing the reward signals and following feasibility rules. Our model represents a parameterized stochastic policy, and by applying a policy gradient algorithm to optimize its parameters, the trained model produces the solution as a sequence of consecutive actions in real time, without the need to re-train for every new problem instance. On capacitated VRP, our approach outperforms classical heuristics and Google's OR-Tools on medium-sized instances in solution quality with comparable computation time (after training). We demonstrate how our approach can handle problems with split delivery and explore the effect of such deliveries on the solution quality. Our proposed framework can be applied to other variants of the VRP such as the stochastic VRP, and has the potential to be applied more generally to combinatorial optimization problems.

cs.AI↗

arXiv2018

Reward Maximization in General Dynamic Matching Systems

We consider a matching system with random arrivals of items of different types. The items wait in queues -- one per each item type -- until they are "matched." Each matching requires certain quantities of items of different types; after a matching is activated, the associated items leave the system. There exists a finite set of possible matchings, each producing a certain amount of "reward". This model has a broad range of important applications, including assemble-to-order systems, Internet advertising, matching web portals, etc. We propose an optimal matching scheme in the sense that it asymptotically maximizes the long-term average matching reward, while keeping the queues stable. The scheme makes matching decisions in a specially constructed virtual system, which in turn control decisions in the physical system. The key feature of the virtual system is that, unlike the physical one, it allows the queues to become negative. The matchings in the virtual system are controlled by an extended version of the greedy primal-dual (GPD) algorithm, which we prove to be asymptotically optimal -- this in turn implies the asymptotic optimality of the entire scheme. The scheme is real-time, at any time it uses simple rules based on the current state of virtual and physical queues. It is very robust in that it does not require any knowledge of the item arrival rates, and automatically adapts to changing rates. The extended GPD algorithm and its asymptotic optimality apply to a quite general queueing network framework, not limited to matching problems, and therefore is of independent interest.

math.PR↗