Sel4FT:用于预训练的注释选择 - - 精细调整与分配转移的微调
概括
我们引入了主动微调,这是选择信息样本标签计算机视觉模型的新方法. 我们的Sel4FT框架有效地识别了各种各样的子集,通过显著的加速度提高了模型性能.
科学领域:
- 计算机视觉 计算机视觉
- 机器学习 机器学习
- 人工智能的人工智能
背景情况:
- 预训练微调模式是计算机视觉中的标准.
- 在微调过程中有效利用有限的注释预算是一个关键的挑战.
- 现有的方法缺乏有效的策略,在预算限制下进行样本选择.
研究的目的:
- 引入主动微调作为一个新的任务,以实现最佳的样本选择.
- 提出Sel4FT,一个统一的框架来选择信息和多样化的数据子集.
- 开发Sel4FT++来处理因数据增强引起的分配转移.
主要方法:
- Sel4FT优化了连续特征空间中的参数模型,以选择子集.
- 选择过程保留了完整的数据池的分布,并保持了多样性.
- Sel4FT++包含了增强意识机制,以应对分配转移.
- 理论分析证明了子集和全池之间的地球移动器距离的最小化.
主要成果:
- 在各种任务中,Sel4FT和Sel4FT++实现了最先进的性能.
- 在图像分类,长尾识别和语义细分方面表现出有效性.
- 与现有的注释选择方法相比,实现了超过100倍的加快速度.
- 在选择过程中消除了代的重新培训和注释.
结论:
- 通过Sel4FT进行主动微调,为现实世界部署提供了高效的解决方案.
- 该框架显著提高了模型性能,同时优化了注释预算.
- Sel4FT为计算机视觉中的数据选择提供了一种强大而可扩展的方法.
相关概念视频
01:25Choosing Between z and t Distribution
2.9K
The z and the Student t distribution estimate the population mean using the sample mean and standard deviation. However, to decide which distribution to use for a calculation, one needs to determine the sample size, the nature of the distribution, and whether the population standard deviation is known. If the population standard deviation is known and the population is normally distributed, or if the sample size is greater than 30, the z distribution is preferred. The Student t distribution is...
2.9K
01:26Distributions to Estimate Population Parameter
4.3K
The accurate values of population parameters such as population proportion, population mean, and population standard deviation (or variance) are usually unknown. These are fixed values that can only be estimated from the data collected from the samples. The estimates of each of these parameters are sample proportion, the sample mean, and sample standard deviation (or variance). To obtain the values of these sample statistics, data are required that have particular distribution and central...
4.3K
01:19Survival Tree
163
Survival trees are a non-parametric method used in survival analysis to model the relationship between a set of covariates and the time until an event of interest occurs, often referred to as the "time-to-event" or "survival time." This method is particularly useful when dealing with censored data, where the event has not occurred for some individuals by the end of the study period, or when the exact time of the event is unknown.
Building a Survival Tree
Constructing a...
Building a Survival Tree
Constructing a...
163
01:12Sampling Distribution
13.6K
Given simple random samples of size n from a given population with a measured characteristic such as mean, proportion, or standard deviation for each sample, the probability distribution of all the measured characteristics is called a sampling distribution. How much the statistic varies from one sample to another is known as the sampling variability of a statistic. You typically measure the sampling variability of a statistic by its standard error. The standard error of the mean is an example...
13.6K
02:07Improving Translational Accuracy
11.9K
Base complementarity between the three base pairs of mRNA codon and the tRNA anticodon is not a failsafe mechanism. Inaccuracies can range from a single mismatch to no correct base pairing at all. The free energy difference between the correct and nearly correct base pairs can be as small as 3 kcal/ mol. With complementarity being the only proofreading step, the estimated error frequency would be one wrong amino acid in every 100 amino acids incorporated. However, error frequencies observed in...
11.9K
01:46Types of Selection
41.5K
Natural selection influences the frequencies of particular alleles and phenotypes within populations in several different ways. Primarily, natural selection can be directional, stabilizing, or disruptive. Directional selection favors one extreme trait and shifts the population towards that phenotype while selecting against individuals displaying alternate traits. Stabilizing selection favors an intermediate trait with a narrow range of variation. Deviation from the optimal phenotype towards an...
41.5K