DeniuHe的博客

Active learning via local structure reconstruction

主动学习论文复现

核逻辑回归

ALC

什么是Expected model change maximization

AL中初始outlier会阻止算法收敛

AL

Ordinal rank is closely related to multiclass classification and metric regression

asymptotic (infinite-training-sample)

Discriminative Locality Alignment (DLA) reduces to Linear Discriminant Analysis (LDA) under specific parameter setting.

Density-weighted uncertainty sampling

生成模型收敛于少量的样本？

Expected Error Reduction的缺点

smoothness assumption

Pang-Ning Tan et al. 2006. Introduction to data mining. Pearson Education India

# ---------------------------------------------------这是一篇基于多视图的主动学习方法。文章提出了一种 Adaptive Maximum Disagreement Criterion# ----------------------------------------------------自适应最大不一致准则的定义如下：原理：一个样本对应K个视图，每个视图对应一个分类器。在当前迭代中，使用当前K个分类器对各视图上的无标记样本进行预...

1. Here, we have to distinguish between the aleatoric uncertainty that is caused by high Bayesian error, and the epistemic uncertainty, which is caused by a lack of information[22].We are not able to reduce the aleatoric uncertainty, but we can acquire m

ESD

1. Active learning algorithm require robust classifier for robust usefulness estimation. Therefore, we choose generative classifiers, namely the Parzen window classifier (PWC)题外As the optimization of the overall performance level is not the scope of th

1. The last strategy belongs to the expected error reduction based methods. The original Value of Information (VoI) criterion as suggested by Joshi et al.[13] selects the instance x_i that minimizes a risk measure defined by them.It has to mentioned that

1. Further，the effectiveness of the proposed approach has been depicted by comparing its performance with state-of-the-art AL algorithms in terms of classification accuracy and computational times as well. The ELM-based AL with different query strategies w

1. Naively labeling all of the data can be prohibitively expensive.

import numpy as npimport pandas as pdfrom scipy.stats import normfrom scipy.stats import multivariate_normal as mvnfrom scipy.stats import multinomialnp.random.seed(231)K = 3#### For diagnostics. Check if covariance matrix is singular.##def i.

1.The authors are grateful to the anonymous reviewers for their comments and suggestions to improve the presentation of this paper.陆续更新

Random sampling的缺点：

1. In many potential applications of machine learning, unlabeled data are abundantly available at low cost, but there is a paucity of labeled data, and labeling unlabeled examples is expensive and / or time-consuming.

He D, Yu H, Wang G, et al. A two-stage clustering-based cold-start method for active learning[J]. Intelligent Data Analysis, 2021, 25(5): 1169-1185.Abstract:The problem of initialization of active learning is considered in this paper. Especially, this p.