检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Bonani, Mayara E. Schwarz, Max Behnke, Sven Computer Science VI University of Bonn Germany Lamarr Institute for Machine Learning and Artificial Intelligence Germany Center for Robotics University of Bonn Germany

Domain adaptation is especially important for robotics applications, where target domain training data is usually scarce and annotations are costly to obtain. We present a method for self-supervised domain adaptation for the scenario where annotated source domain data (e.g. from synthetic generation) is available, but the target domain data is completely unannotated. Our method targets the semantic segmentation task and leverages a segmentation foundation model (Segment Anything Model) to obtain segment information on unannotated data. We take inspiration from recent advances in unsupervised local feature learning and propose an invariance-variance loss over the detected segments for regularizing feature representations in the target domain. Crucially, this loss structure and network architecture can handle overlapping segments and oversegmentation as produced by Segment Anything. We demonstrate the advantage of our method on the challenging YCB-Video and HomebrewedDB datasets and show that it outperforms prior work and, on YCB-Video, even a network trained with real annotations. Additionally, we provide insight through model ablations and show applicability to a custom robotic application. Copyright © 2023, The Authors. All rights reserved.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

Forecasting of Traffic Flow Using Feature Selection with ML Model

Forecasting of Traffic Flow Using Feature Selection with ML ...

引用

2023 International Conference on Data science and Network Security, ICDSNS 2023

作者： Kumar Pareek, Piyush Vidyalakshmi, K. Nayana, S. Gangambika, G. Deepak, H.A. Nitte Meenakshi Institute of Technology Department of Artificial Intelligence and Machine Learning Bengaluru India Sri Siddhartha Institute of Technology Department of Information Science and Engineering Tumakuru India East West Institute of Technology Department of Computer Science and Engineering Bengaluru India Navkis College of Engineering Department of Electronics and Communication Engineering Hassan India

ISBN: (纸本)9798350301595

Intelligent traffic control and public risk assessment are only two examples of the many spatial-temporal mining applications that rely on accurate predictions of citywide traffic flow. The ability to select an optimal traffic variable grouping is crucial to traffic flow prognostication, as it can cut down on calculation time and boost accuracy. Although the linear traffic system problem can be easily solved using the conventional method, its inability to adapt to sudden shifts in traffic makes it challenging to use. The intelligent forecasting approach works well for managing short-term traffic flow and can realise arbitrary non-linear mapping. Unfortunately, its effectiveness is highly sensitive to the supplied data. Parameter selection is another challenging topic that necessitates a big training set. To this end, this work proposes a feature selection method employing a hybrid optimization model (the slime procedure (SMA) and the whale optimization-(WOA)). To begin, the suggested model assesses the importance and uniqueness of each variable, and then uses feature selection to pick the most important ones and eliminate the duplicates. Secondly, the forecasting engine is Extreme machine learning (ELM). Lastly, the suggested technique's forecasting performance is validated using a numerical instance using traffic flow data from Pems;the consequences show that the proposed method may successfully lower the cost of control while simultaneously increasing the model's prognostication precision. © 2023 IEEE.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

Breast Cancer Classification Model Using Principal Component Analysis and Deep Neural Network 1

引用

World Conference on Information Systems for Business Management, ISBM 2022

作者： Sindhuja, M. Poonkuzhali, S. Vigneshwaran, P. Department of Artificial Intelligence and Machine Learning Rajalakshmi Engineering College Tamil Nadu Thandalam Chennai602105 India Department of Computer Science and Engineering Rajalakshmi Engineering College Tamil Nadu Thandalam Chennai602105 India Department of Networking and Communications SRM Institute of Science and Technology Tamil Nadu Kattankulathur Chengalpattu603203 India

ISBN: (数字)9789811974472

ISBN: (纸本)9789811974465

Life-threatening cancer is prevalent over the globe. According to statistics, most people are diagnosed with cancer in the later stages, even though cancer can be prevented and cured in early stages. The goal of this research is to diagnose a breast cancer is either benign or malignant, as well as to forecast the likelihood of a cancer recurrence even after a course of therapy has been completed. Despite the fact that many machine learning algorithms have resulted in strong predictions, the accuracy in the early stages of categorization is not up to the expected level. Deep learning (DL), a higher degree of machine learning, can forecast breast cancer types and recurrences. Classifiers were built using a deep neural network (DNN) that used Principal Component Analysis (PCA) to choose features. Different machine learning techniques are compared with the proposed system’s accuracy. Early-stage breast cancer prediction is more accurate with the DNN-based method. The clinical management system will benefit from the proposed system since it will aid in identification of cancer at an early stage and the subsequent provision of appropriate therapy. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

StratXplore: Strategic Novelty-seeking and Instruction-aligned Exploration for Vision and Language Navigation

arXiv

引用

arXiv 2024年

作者： Gopinathan, Muraleekrishna Abu-Khalaf, Jumana Suter, David Masek, Martin Centre for Artificial Intelligence and Machine Learning School of Science Edith Cowan University 270 Joondalup Dr JoondalupWA6027 Australia

Embodied navigation requires robots to understand and interact with the environment based on given tasks. Vision-Language Navigation (VLN) is an embodied navigation task, where a robot navigates within a previously seen and unseen environment, based on linguistic instruction and visual inputs. VLN agents need access to both local and global action spaces;former for immediate decision making and the latter for recovering from navigational mistakes. Prior VLN agents rely only on instruction-viewpoint alignment for local and global decision making and back-track to a previously visited viewpoint, if the instruction and its current viewpoint mismatches. These methods are prone to mistakes, due to the complexity of the instruction and partial observability of the environment. We posit that, back-tracking is suboptimal and agent that is aware of its mistakes can recover efficiently. For optimal recovery, exploration should be extended to unexplored viewpoints (or frontiers). The optimal frontier is a recently observed but unexplored viewpoint that aligns with the instruction and is novel. We introduce a memory-based and mistake-aware path planning strategy for VLN agents, called StratXplore, that presents global and local action planning to select the optimal frontier for path correction. The proposed method collects all past actions and viewpoint features during navigation and then selects the optimal frontier suitable for recovery. Experimental results show this simple yet effective strategy improves the success rate on two VLN datasets with different task complexities. © 2024, CC BY.

关键词： Visual languages

来源：评论

学校读者我要写书评

暂无评论

Emotion Recognition using Deep Stacked Autoencoder with Softmax Classifier

Emotion Recognition using Deep Stacked Autoencoder with Soft...

引用

artificial intelligence and Smart Systems (ICAIS), International Conference on

作者： M. Mohana P. Subashini Computer Science Centre for Machine Learning and Intelligence (CMLI) Avinashilingam Institute Coimbatore India

Deep learning and computer vision research are still quite active in the field of facial emotion recognition (FER). It has been widely applied in several research areas but not limited to human-robot interaction, human psychology interaction detection, and learners’ emotion identification. In recent decades, facial expression recognition using deep learning has proven to be effective. This performance has been achieved by a good degree of self-learn kernels in the convolution layer which retains spatial information of images with higher accuracy. Even though, it often leads to convergence in non-optimal local minima due to randomized initialization of weights. This paper introduces a Deep stacked autoencoder in which the output of one autoencoder has given into the input of another autoencoder along with input values. A single autoencoder does not sufficient to extract the complex relationship in features. So, these concatenated features of the stacked autoencoder help to focus on highly active features during training and testing. In addition, this approach helps to solve inefficient data issues. Finally, trained autoencoders have fine-tuned with the Adam optimizer, and emotions are classified by a softmax layer. The outcomes of the proposed methodology on the JAFFE dataset are significant, according to experiments. The proposed method achieved 82% of accuracy, 85% of Precision, 82% of Recall, and 81% of F1-score. Additionally, the performance of the stacked autoencoder has been examined using the reconstruction loss and roc curve.

关键词： Training Deep learning Emotion recognition Face recognition Psychology Human-robot interaction Feature extraction

来源：评论

学校读者我要写书评

暂无评论

The Accurate Prediction and Forecasting Strategy Implementation for Various Weather Conditions using Different Algorithms 4

The Accurate Prediction and Forecasting Strategy Implementat...

引用

4th IEEE International Conference on Advance Computing and Innovative Technologies in Engineering, ICACITE 2024

作者： Sivapriyan, R. Vidya Prabha, C.V. Santhakumar, D. Rao, Kolli Himantha Mohammed, Safaa Hallol Abdtawfeeq, Tareq Hafdhi Sir Mvit R & D Centre Department of Eee Bengaluru India Karpagam College of Engineering Department of Computer Science and Engineering Coimbatore India Saveetha Institute of Medical and Technical Sciences Saveetha School of Engineering Department of Computer Science Engineering Chennai India Institute of Cse Saveetha Engineering College Department of Artificial Intelligence and Machine Learning India National University of Science and Technology Collage of Pharmacy Dhi Qar64001 Iraq Al-Farahidi University Baghdad Iraq

ISBN: (纸本)9798350360165

this study presents a new regression method for forecasting the occurrence of rainfall. The method includes three main steps: feature extraction, classification, and preprocessing. In the feature extraction step, all weather samples are normalized for a uniform sampling frequency, after which signal noise is removed to improve data quality and each signal is then converted to a 1 × 34 feature vector, capturing subtle emotional nuances. Notably, normalization precedes classifier training to improve model performance. Performance metrics including f-score, sensitivity, precision, and accuracy are used to evaluate the efficiency of the model. The proposed methodology has the potential to advance our understanding of emotional responses to music, providing a promising avenue for future research in predicting rainfall and other natural phenomenon. © 2024 IEEE.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

Flowmind2Digital: The First Comprehensive Flowmind Recognition and Conversion Approach

arXiv

引用

arXiv 2024年

作者： Liu, Huanyu Cai, Jianfeng Zhang, Tingjia Li, Hongsheng Wang, Siyuan Zhu, Guangming Shah, Syed Afaq Ali Bennamoun, Mohammed Zhang, Liang School of Artificial Intelligence Xidian University China School of Computer Science and Technology Xidian University China School of Science Centre for AI and Machine Learning ECU School of Physics Maths and Computing Computer Science and Software Engineering UWA

Flowcharts and mind maps, collectively referred to as flowmind, play an important role in our daily, and many companies have developed dedicated tools. Hand-drawn flowminds offer significant advantages for real-time and collaborative communication. However, there is an increasing demand to convert them into a digital format for further processing. Automated conversion methods are crucial in addressing the challenges associated with manual conversion, such as the cost of time and learning. Previous works have proposed diverse sketch recognition methods. However, these methods face significant limitations in practical situations. Firstly, most methods are designed for specific fields, making it challenging to extend their usage to other fields. Additionally, none of these methods address the critical step of digital conversion after recognition, which is essential to users. Moreover, existing datasets exhibit significant biases relative to actual data, hindering the methods’ generalization ability. Our paper proposes the Flowmind2digital method and hdFlowmind dataset to address the aforementioned challenges. Flowmind2digital is the first comprehensive recognition and conversion method for flowminds, utilizing a neural network architecture and keypoint detection technology to enhance overall recognition accuracy. Our hdFlowmind dataset consists of 1,776 hand-drawn and manually annotated flowminds, covering 22 scenarios and surpassing existing datasets in size. Our experiments showcase the effectiveness of our method, with an accuracy rate of 87.3% on the hdFlowmind dataset, surpassing the previous state-of-the-art work by 11.9%. Additionally, our dataset demonstrates effectiveness, with a 2.9% increase in accuracy after pre-training and fine-tuning on Handwritten-diagram-dataset. We also highlight the importance of simple graphics for sketch recognition, which can improve accuracy by 9.3%. Copyright © 2024, The Authors. All rights reserved.

关键词： Image recognition

来源：评论

学校读者我要写书评

暂无评论

Unsupervised Pre-Training for 3D Leaf Instance Segmentation

arXiv

引用

arXiv 2024年

作者： Roggiolani, Gianmarco Magistri, Federico Guadagnino, Tiziano Behley, Jens Stachniss, Cyrill University of Bonn Germany Department of Engineering Science the University of Oxford United Kingdom Lamarr Institute for Machine Learning and Artificial Intelligence Germany

Crops for food, feed, ber, and fuel are key natural resources for our society. Monitoring plants and measuring their traits is an important task in agriculture often referred to as plant phenotyping. Traditionally, this task is done manually, which is time- and labor-intensive. Robots can automate phenotyping providing reproducible and high-frequency measurements. Today’s perception systems use deep learning to interpret these measurements, but require a substantial amount of annotated data to work well. Obtaining such labels is challenging as it often requires background knowledge on the side of the labelers. This paper addresses the problem of reducing the labeling effort required to perform leaf instance segmentation on 3D point clouds, which is a rst step toward phenotyping in 3D. Separating all leaves allows us to count them and compute relevant traits as their areas, lengths, and widths. We propose a novel self-supervised task-specic pre-training approach to initialize the backbone of a network for leaf instance segmentation. We also introduce a novel automatic postprocessing that considers the difculty of correctly segmenting the points close to the stem, where all the leaves petiole overlap. The experiments presented in this paper suggest that our approach boosts the performance over all the investigated scenarios. We also evaluate the embeddings to assess the quality of the fully unsupervised approach and see a higher performance of our domain-specic postprocessing. © 2024, CC BY.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Your Transformer May Not be as Powerful as You Expect 36

Your Transformer May Not be as Powerful as You Expect

引用

36th Conference on Neural Information Processing Systems, NeurIPS 2022

作者： Luo, Shengjie Li, Shanda Zheng, Shuxin Liu, Tie-Yan Wang, Liwei He, Di National Key Laboratory of General Artificial Intelligence School of Intelligence Science and Technology Peking University China Machine Learning Department School of Computer Science Carnegie Mellon University United States Microsoft Research United States Center for Data Science Peking University China Zhejiang Lab China

ISBN: (纸本)9781713871088

Relative Positional Encoding (RPE), which encodes the relative distance between any pair of tokens, is one of the most successful modifications to the original Transformer. As far as we know, theoretical understanding of the RPE-based Transformers is largely unexplored. In this work, we mathematically analyze the power of RPE-based Transformers regarding whether the model is capable of approximating any continuous sequence-to-sequence functions. One may naturally assume the answer is in the affirmative-RPE-based Transformers are universal function approximators. However, we present a negative result by showing there exist continuous sequence-to-sequence functions that RPE-based Transformers cannot approximate no matter how deep and wide the neural network is. One key reason lies in that most RPEs are placed in the softmax attention that always generates a right stochastic matrix. This restricts the network from capturing positional information in the RPEs and limits its capacity. To overcome the problem and make the model more powerful, we first present sufficient conditions for RPE-based Transformers to achieve universal function approximation. With the theoretical guidance, we develop a novel attention module, called Universal RPE-based (URPE) Attention, which satisfies the conditions. Therefore, the corresponding URPE-based Transformers become universal function approximators. Extensive experiments covering typical architectures and tasks demonstrate that our model is parameter-efficient and can achieve superior performance to strong baselines in a wide range of applications. The code will be made publicly available at https://***/lsj2408/URPE. © 2022 Neural information processing systems foundation. All rights reserved.

关键词： Encoding (symbols)

来源：评论

学校读者我要写书评

暂无评论

Sample complexity of posted pricing for a single item 24

Sample complexity of posted pricing for a single item

引用

Proceedings of the 38th International Conference on Neural Information Processing Systems

作者： Billy Jin Thomas Kesselheim Will Ma Sahil Singla Cornell University School of Operations Research and Information Engineering University of Bonn Institute of Computer Science and Lamarr Institute for Machine Learning and Artificial Intelligence Columbia University Graduate School of Business and Data Science Institute Georgia Tech School of Computer Science

ISBN: (纸本)9798331314385

Selling a single item to n self-interested buyers is a fundamental problem in economics, where the two objectives typically considered are welfare maximization and revenue maximization. Since the optimal mechanisms are often impractical and do not work for sequential buyers, posted pricing mechanisms, where fixed prices are set for the item for different buyers, have emerged as a practical and effective alternative. This paper investigates how many samples are needed from buyers' value distributions to find near-optimal posted prices, considering both independent and correlated buyer distributions, and welfare versus revenue maximization. We obtain matching upper and lower bounds (up to logarithmic factors) on the sample complexity for all these settings.

关键词：

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：