<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[Xie_learning的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/Xie_learning</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; Xie_learning]]></copyright><item><title><![CDATA[Leveraging the Invariant Side of Generative Zero-Shot Learning【CVPR2019】]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/104017830</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/104017830</guid><author>Xie_learning</author><pubDate>Fri, 17 Jan 2020 15:25:02 +0800</pubDate><description><![CDATA[PDF:Leveraging the Invariant Side of Generative Zero-Shot Learning
code:implement by pytorch
摘要
Conventional zero-shot learning (ZSL) methods generally learn an embedding, e.g., visual-semantic mappin...]]></description><category></category></item><item><title><![CDATA[Generalized Zero- and Few-Shot Learning via Aligned Variational Autoencoders【CVPR2019】]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/99679265</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/99679265</guid><author>Xie_learning</author><pubDate>Fri, 16 Aug 2019 17:02:24 +0800</pubDate><description><![CDATA[Generalized Zero- and Few-Shot Learning via Aligned Variational Autoencoders摘要网络结构图
![在这里插入图片描述](https://img-blog.csdnimg.cn/20190816160725575.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdG...]]></description><category></category></item><item><title><![CDATA[Meta-learning with latent embedding optimization【论文笔记】]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/96150265</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/96150265</guid><author>Xie_learning</author><pubDate>Wed, 17 Jul 2019 11:33:17 +0800</pubDate><description><![CDATA[Gradient-based meta-learning techniques are both widely applicable and proficient at solving challenging few-shot learning and fast adaptation problems. However, they have practical difficulties when operating on high-dimensional parameter spaces in extrem]]></description><category></category></item><item><title><![CDATA[deepin15.9安装PyTorch,cuda和cudnn]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/90735409</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/90735409</guid><author>Xie_learning</author><pubDate>Sun, 02 Jun 2019 02:57:46 +0800</pubDate><description><![CDATA[deepin安装pytorch,cuda,cudnn
本文针对deepin15.9

进入Bios关闭集成显卡
对deepin换源

第一步：备份官方默认源
sudo cp /etc/apt/sources.list /etc/apt/sources.list.bak
第二步：打开sources.list
sudo gedit /etc/apt/sources.list
第三步：换源
## Gen...]]></description><category></category></item><item><title><![CDATA[步长为s的二维卷积Python实现]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/90257750</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/90257750</guid><author>Xie_learning</author><pubDate>Thu, 16 May 2019 10:11:24 +0800</pubDate><description><![CDATA[import numpy as np

x = np.arange(1, 21).reshape(5, 4)
w = np.array([[1, 1], [-1, -1]])

def my_conv(input, kernel, s):
    output_size_0 = int((len(input) - len(kernel)) / s + 1)   # 输出结果的第0维长度
    o...]]></description><category></category></item><item><title><![CDATA[将分类存储的图片切分为训练集、验证集和测试集(PyTorch实现）]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/89646101</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/89646101</guid><author>Xie_learning</author><pubDate>Sun, 28 Apr 2019 21:14:55 +0800</pubDate><description><![CDATA[在实验中，经常遇到图片数据是分类存储的—即不同的文件夹存放不同label的图片，比如猫狗数据集，将猫的数据放在cat文件夹下，而将狗的图片放在名为dog的文件夹。在跑模型的过程中，常常需要将数据切分为训练集、验证集和测试集，并且同一类别的label要一致。
from torchvision.datasets import ImageFolder
from PIL import Image
from...]]></description><category></category></item><item><title><![CDATA[PyTorch将CIFAR100数据按类标归类保存]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/89365305</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/89365305</guid><author>Xie_learning</author><pubDate>Wed, 17 Apr 2019 21:27:09 +0800</pubDate><description><![CDATA[few-shot learning的采样
Few-shot learning 基于任务对模型进行训练，在N-way-K-shot中，一个任务中的meta-training中含有N类,每一类抽取K个样本构成support set, query set则是在刚才抽取的N类剩余的样本中sample一定数量的样本（可以是均匀采样，也可以是不均匀采样）。
对数据按类标归类
针对上述情况，我们需要使用不同类别...]]></description><category></category></item><item><title><![CDATA[Python创建并保存json文件，支持数据更新保存]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/89176812</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/89176812</guid><author>Xie_learning</author><pubDate>Wed, 10 Apr 2019 10:58:53 +0800</pubDate><description><![CDATA[import json

class Params():
    """Class that loads hyperparameters from a json file.
        Example:
        ```
        params = Params(json_path)
        print(params.learning_rate)
        param...]]></description><category></category></item><item><title><![CDATA[PyTorch加载预训练模型(pretrained)]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/89176636</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/89176636</guid><author>Xie_learning</author><pubDate>Wed, 10 Apr 2019 10:53:40 +0800</pubDate><description><![CDATA[使用预训练模型的代码如下：
# 加载预训练模型
    resNet50 = models.resnet50(pretrained=True)
    ResNet50 = ResNet(Bottleneck, [3, 4, 6, 3], num_classes=2)

    # 读取参数
    pretrained_dict = resNet50.state_dict()
    model...]]></description><category></category></item><item><title><![CDATA[transforms图片处理操作]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/88911373</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/88911373</guid><author>Xie_learning</author><pubDate>Sat, 30 Mar 2019 13:45:53 +0800</pubDate><description><![CDATA[使用jupyter notebook编译器
from PIL import Image
from torchvision import transforms as T
import torch as t

to_tensor = T.ToTensor()
to_pil = T.ToPILImage()
lena = Image.open('/home/xie/图片/lena.jpg')

lena...]]></description><category></category></item><item><title><![CDATA[神经网络中的交叉熵]]></title><link>https://blog.youkuaiyun.com/Xie_learning/article/details/88732845</link><guid>https://blog.youkuaiyun.com/Xie_learning/article/details/88732845</guid><author>Xie_learning</author><pubDate>Fri, 22 Mar 2019 10:44:03 +0800</pubDate><description><![CDATA[神经网络中的交叉熵作为分类问题中的损失函数PyTorch中的交叉熵
作为分类问题中的损失函数
在神经网络的多分类问题中，常常使用one-hot的方法设置n类输出，比如在手写体识别问题中，如果是数字1，神经网络的输出越接近[0,1,0,0,0,0,0,0,0,0]越好。那么如何衡量神经网络与label的接近程度呢？交叉熵是常用的使用方法。其刻画的是两个概率分布之间的距离。
交叉熵是信息论中的概念。对...]]></description><category></category></item></channel></rss>