Multi-Agent Image Classification via Reinforcement Learning

Mousavi, Hossein; Nazari, Mohammadreza; Takáč, Martin; Motee, Nader

doi:10.1109/iros40897.2019.8968129

Cited by 16 publications

(11 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to acquire adequate observations, the robot could consider traversing the environment while collecting local information as sequential data and attempting to learn their inter-correlation with recurrent neural networks. The same type of framework that uses localized observations and recurrent neural networks to learn the image and map classification is illustrated in our previous works [22,21,16,17], and shows promising performance in various scenarios.…”

Section: Introductionmentioning

confidence: 85%

“…In the first case study, we consider the robot is traveling over the image from the MNIST dataset [14,22]. The training and testing is performed in PyTorch [23] ADAM [10] and a learning rate l r = 0.0001.…”

Section: B Mnist Datasetmentioning

confidence: 99%

“…It is shown that a single robot can classify by only revealing a small portion of the environment. It should be emphasized that the performance on the MNIST will reach 98% by using multiple communicating robots [22].…”

Section: B Mnist Datasetmentioning

confidence: 99%

See 2 more Smart Citations

Robustness Analysis of Classification Using Recurrent Neural Networks with Perturbed Sequential Input

Amini¹,

Takáč²,

Motee³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

For a given stable recurrent neural network (RNN) that is trained to perform a classification task using sequential inputs, we quantify explicit robustness bounds as a function of trainable weight matrices. The sequential inputs can be perturbed in various ways, e.g., streaming images can be deformed due to robot motion or imperfect camera lens. Using the notion of the Voronoi diagram and Lipschitz properties of stable RNNs, we provide a thorough analysis and characterize the maximum allowable perturbations while guaranteeing the full accuracy of the classification task. We illustrate and validate our theoretical results using a map dataset with clouds as well as the MNIST dataset.

show abstract

Section: Introductionmentioning

confidence: 85%

Section: B Mnist Datasetmentioning

confidence: 99%

See 1 more Smart Citation

Robustness Analysis of Classification Using Recurrent Neural Networks with Perturbed Sequential Input

Amini¹,

Takáč²,

Motee³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…A common model for RL is the standard Markov Decision Process. , RL can be divided into model-based RL and model-free RL, , as well as active RL and passive RL . DL models can also be used in RL to form deep RL (DRL). , These methods have been applied to diverse fields in biology and chemistry, including drug discovery, , protein design, , and chemical engineering. , They have also been used in image recognition , and financial markets . Next, we will summarize how RL can be applied to the study of biochemical molecules in the context of small data sets.…”

Section: Methods For Small Molecular Data Challengesmentioning

confidence: 99%

Machine Learning Methods for Small Data Challenges in Molecular Science

et al. 2023

View full text Add to dashboard Cite

Small data are often used in scientific and engineering research due to the presence of various constraints, such as time, cost, ethics, privacy, security, and technical limitations in data acquisition. However, big data have been the focus for the past decade, small data and their challenges have received little attention, even though they are technically more severe in machine learning (ML) and deep learning (DL) studies. Overall, the small data challenge is often compounded by issues, such as data diversity, imputation, noise, imbalance, and high-dimensionality. Fortunately, the current big data era is characterized by technological breakthroughs in ML, DL, and artificial intelligence (AI), which enable data-driven scientific discovery, and many advanced ML and DL technologies developed for big data have inadvertently provided solutions for small data problems. As a result, significant progress has been made in ML and DL for small data challenges in the past decade. In this review, we summarize and analyze several emerging potential solutions to small data challenges in molecular science, including chemical and biological sciences. We review both basic machine learning algorithms, such as linear regression, logistic regression (LR), k-nearest neighbor (KNN), support vector machine (SVM), kernel learning (KL), random forest (RF), and gradient boosting trees (GBT), and more advanced techniques, including artificial neural network (ANN), convolutional neural network (CNN), U-Net, graph neural network (GNN), Generative Adversarial Network (GAN), long short-term memory (LSTM), autoencoder, transformer, transfer learning, active learning, graph-based semi-supervised learning, combining deep learning with traditional machine learning, and physical model-based data augmentation. We also briefly discuss the latest advances in these methods. Finally, we conclude the survey with a discussion of promising trends in small data challenges in molecular science.

show abstract

“…[ 11 ], where each agent controls a sliding square window. This method is more robust and has better generalization based on the results of the experiments on the CIFAR-10 dataset and the MNIST dataset [ 21 ].…”

Section: Introductionmentioning

confidence: 99%

Image Classification Method Based on Multi-Agent Reinforcement Learning for Defects Detection for Casting

Liu

Zhang

Mao

2022

Sensors

View full text Add to dashboard Cite

A casting image classification method based on multi-agent reinforcement learning is proposed in this paper to solve the problem of casting defects detection. To reduce the detection time, each agent observes only a small part of the image and can move freely on the image to judge the result together. In the proposed method, the convolutional neural network is used to extract the local observation features, and the hidden state of the gated recurrent unit is used for message transmission between different agents. Each agent acts in a decentralized manner based on its own observations. All agents work together to determine the image type and update the parameters of the models by the stochastic gradient descent method. The new method maintains high accuracy. Meanwhile, the computational time can be significantly reduced to only one fifth of that of the GhostNet.

show abstract

Multi-Agent Image Classification via Reinforcement Learning

Cited by 16 publications

References 16 publications

Robustness Analysis of Classification Using Recurrent Neural Networks with Perturbed Sequential Input

Robustness Analysis of Classification Using Recurrent Neural Networks with Perturbed Sequential Input

Machine Learning Methods for Small Data Challenges in Molecular Science

Image Classification Method Based on Multi-Agent Reinforcement Learning for Defects Detection for Casting

Contact Info

Product

Resources

About