- 博客(10)
- 收藏
- 关注
原创 K近邻分类模型
输入:N个特征向量:T={(x1,y1),(x2,y2),...,(xN,yN)}T = \lbrace (x_1, y_1), (x_2, y_2), ..., (x_N, y_N) \rbraceT={(x1,y1),(x2,y2),...,(xN,yN)} ,xi∈Rnx_i \in R^nxi∈Rn, i.e. xi=(x(1),x(2),...,x(n))x_i = (x^{(1)}, x^{(2)}, ..., x^{(n)})xi=(x(1),x(2),...,x(n))输.
2022-05-07 17:44:46
185
原创 水塘抽样(Reservoir Sampling)
水塘抽样(Reservoir Sampling): 力扣398题目而来问题描述:当内存无法加载全部数据时,如何从包含未知大小的数据流中随机选取k个数据,并且要保证每个数据被抽取到的概率相等。特殊且常用情况:k=1,假设数据量为N第一个数n1n_1n1,我们选择保留,p(n1)=1p(n_1) = 1p(n1)=1第二个数n2n_2n2,我们以12\frac{1}{2}21的概率保留,那么n1n_1n1被保留(不被覆盖)的概率p(n1)=1∗(1−12)=12p(n_1) = 1*(.
2022-04-26 10:52:35
1886
原创 Python细枝末节个人知识记录
from x import * 可以选择哪些不被引入不能引入x.py中前面带下划线的变量,如:_y=10可以定义 __all__变量指示哪些变量可以引入,如 __all__ =[‘a’, ‘lt’]dir(module) 可以显示module对象所有的方法如果类中的变量不想让用户直接修改,可以将变量定义为内部属性,如:class account: def __init__(self, name): self.__name = name def rename(self, name.
2022-04-04 19:50:09
695
原创 深度学习优化器个人经验汇总
通用公式gt=▽f(wt)根据loss计算出的梯度值mt=ϕ(g1,g2,...,gt)根据过去梯度惯性和当前梯度算出的一阶动量Vt=φ(g1,g2,...,gt)根据过去梯度和当前梯度算出的二阶动量ηt=αVt∗mt根据一阶和二阶动量以及学习率α计算需要更新的梯度wt+1=wt−ηt更新梯度g_t = \triangledown f(w_t) \qquad \tiny{根据loss计算出的梯度值} \\\normalsize{m_t = \phi (g_1, g_2, ..., g_t)} \.
2022-04-03 21:04:43
2657
原创 阅读笔记:Universal Domain Adaptation through Self-Supervision
paper: https://arxiv.org/pdf/2002.07953.pdfcode: https://github.com/VisionLearningGroup/DANCE参考请引用:Saito, K., Kim, D., Sclaroff, S., & Saenko, K. (2020). Universal Domain Adaptation through Self Supervision. ArXiv, abs/2002.07953.Motivation:当前大部
2020-12-19 21:57:44
1438
4
原创 阅读笔记:Structured Domain Adaptation with Online Relation Regularization for Unsupervised Person Re-ID
paper: https://arxiv.org/pdf/2003.06650.pdfcode: https://github.com/yxgeee/SDA参考请引用:Ge, Y., Zhu, F., Zhao, R., & Li, H. (2020). Structured Domain Adaptation with Online Relation Regularization for Unsupervised Person Re-ID. arXiv: Computer Vision
2020-12-19 21:30:52
360
1
原创 阅读笔记:Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID
阅读笔记:Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-IDpaper: https://arxiv.org/pdf/2006.02713.pdfcode: Github: https://github.com/yxgeee/SpCL参考请引用:Ge, Y., Chen, D., Zhu, F., Zhao, R., & Li, H. (2020). Self-paced C
2020-12-19 21:10:38
333
3
原创 文章阅读:UNSUPERVISED REPRESENTATION LEARNING BY PREDICTING RANDOM DISTANCES
UNSUPERVISED REPRESENTATION LEARNING BY PREDICTING RANDOM DISTANCES文章:https://arxiv.org/pdf/1912.12186.pdf代码:https://git.io/RDP参考请引用:Wang, H., Pang, G., Shen, C., & Ma, C. (2019). Unsupervised Representation Learning by Predicting Random Distances
2020-12-19 16:48:40
912
3
空空如也
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人