Adam婷
笔者在人工智能/机器学习领域中默默探索,时而迷惘,时而欣喜。
展开
-
使用OpenCV对物体搜索检测与识别
在本教程中,我们将了解对象检测中称为“选择性搜索”的重要概念。我们还将用C ++和Python共享OpenCV代码。物体检测与物体识别对象识别算法识别图像中存在哪些对象。它将整个图像作为输入,并输出该图像中存在的对象的类标签和类概率。例如,类标签可以是“狗”,相关的类概率可以是97%。另一方面,对象检测算法不仅告诉您图像中存在哪些对象,还输出边界框(x,y,宽度,高度)以指示图像内对象的位置。所有物体检测算法的核心是物体识别算法。假设我们训练了一个物体识别模型,该模型识别图像斑块中的狗。该模型将判断原创 2020-09-22 17:14:43 · 11643 阅读 · 5 评论 -
Cascade Classifier Training Using OpenCV
IntroductionWorking with a boosted cascade of weak classifiers includes two major stages: the training and the detection stage. The detection stage using either HAAR or LBP based models, is described in the object detection tutorial. This documentation gi翻译 2020-09-18 17:46:30 · 369 阅读 · 0 评论 -
Computing Visual Perception to Enable Real-Time Object Recognition(计算视觉感知以启用实时对象识别)
Visual search has become a necessity for many multimedia applications running on today’s computing systems. Tasks such as recognizing parts of a scene in an image, detecting items in a retail store, navigating an autonomous drone, etc. have great relevance翻译 2020-09-17 18:14:23 · 493 阅读 · 0 评论 -
利用OPENCV创作梵高艺术风格图片
OpenCV is a library with 20 years of continuous development under its belt. The age of introspection and search of destiny. Are there any projects based on the library which have made someone’s life better and happier? Can you make it by yourself? Seeking翻译 2020-09-17 18:06:25 · 678 阅读 · 0 评论 -
CornerNet-Lite: Efficient Keypoint-Based Object Detection(CornerNet-Lite:基于关键点的高效对象检测)
As was mentioned before, the good thing about CornerNet is its competitive results on MS COCO dataset. Nevertheless, it has one huge drawback.It is slow.To overcome this issue, the authors proposed CornerNet-Lite – a combination of two efficient varian原创 2020-09-17 17:15:28 · 442 阅读 · 0 评论 -
Understanding Transductive Few-shot Learning
Image annotation can be a bottleneck for the applicability of machine learning when a costly expert annotation is needed. Some examples are medical imaging, astronomy, or botanics.To alleviate this problem, few-shot classification aims to train classifier原创 2020-09-17 10:41:26 · 1178 阅读 · 0 评论 -
Large Scale Evolving Graphs with Burst Detection
Large Scale Evolving Graphs with Burst DetectionAbstractAnalyzing large-scale evolving graphs are crucialfor understanding the dynamic and evolutionary nature of social networks. Most existing works focus on discovering repeated and consistent temporal原创 2020-07-23 22:39:30 · 767 阅读 · 0 评论 -
Controllable Multi-Interest Framework for Recommendation
Controllable Multi-Interest Framework for RecommendationABSTRACTRecently, neural networks have been widely used in e-commerce recommender systems, owing to the rapid development of deep learning. We formalize the recommender system as a sequential recom原创 2020-07-23 21:54:26 · 3477 阅读 · 0 评论 -
ArnetMiner: Extraction and Mining of Academic Social Networks
ArnetMiner: Extraction and Mining of Academic Social NetworksABSTRACTThis paper addresses several key issues in the ArnetMiner system, which aims at extracting and mining academic social networks. Specifically, the system focuses on: 1) Extracting resear原创 2020-07-23 20:23:52 · 3023 阅读 · 0 评论 -
Distributed Stochastic Gradient Method for Non-Convex Problems with Applications in Supervised Learn
Distributed Stochastic Gradient Method for Non-Convex Problems with Applications in Supervised LearningAbstractWe develop a distributed stochastic gradient descent algorithm for solving non-convex optimization problems under the assumption that the local原创 2020-07-22 16:35:00 · 753 阅读 · 0 评论 -
Computing Visual Perception to Enable Real-Time Object Recognition(计算视觉感知以启用实时对象识别)
Computing Visual Perception to Enable Real-Time Object RecognitionReferences:Visual search has become a necessity for many multimedia applications running on today’s computing systems. Tasks such as recognizing parts of a scene in an image, detecting item翻译 2020-07-16 20:56:33 · 568 阅读 · 0 评论 -
强化学习提纲(reinforcement learning notebook)
强化学习(reinforcement learning)被认为是人类通往通用人工智能(ANI)的有效途径。从以Deepmind为代表的研发团队对强化学习在游戏博弈中的突出表现来看,强化学习的无监督的学习方法所展现的效果惊人。强化学习的特点不断试错,也就是通过尝试与环境的交互,来解决策略评估的问题。强化学习的精髓在于“算”,即:通过大量数据的迭代运算,从大量的运算结果样本集中选取最优策略,从而...原创 2019-02-02 20:18:23 · 383 阅读 · 0 评论 -
最近邻居算法精解
经典K-NN算法解析k近邻算法是一种基本分类和回归方法1.近邻算法,即是给定一个训练数据集,对新的输入实例,在训练数据集中找到与该实例最邻近的K个实例,这K个实例的多数属于某个类,就把该输入实例分类到这个类中。(这就类似于现实生活中少数服从多数的思想)更通俗说一遍算法的过程,来了一个新的输入实例,我们算出该实例与每一个训练点的距离(这里的复杂度为0(n)比较大,所以引出了下文的kd树等结构),...原创 2019-02-03 09:06:09 · 1769 阅读 · 0 评论 -
逻辑回归精解
机器学习经典算法----逻辑回归本博客以代码算法为依据重在分析逻辑回归算法的实用性,数据预处理部分视为前期工作。概述逻辑回归(也称“对数几率回归”)(英语:Logistic regression 或logit regression),即逻辑模型(英语:Logit model,也译作“评定模型”、“分类评定模型”)是离散选择法模型之一,属于多重变量分析范畴,是社会学、生物统计学、临床、数量心...原创 2019-02-03 11:05:33 · 222 阅读 · 0 评论 -
一元线性回归与多元线性回归
线性回归action精讲线性回归是利用数理统计中回归分析,来确定两种或两种以上变量间相互依赖的定量关系的一种统计分析方法,运用十分广泛。其表达形式为y = w’x+e,e为误差服从均值为0的正态分布。一元与多元回归分析中,只包括一个自变量和一个因变量,且二者的关系可用一条直线近似表示,这种回归分析称为一元线性回归分析。如果回归分析中包括两个或两个以上的自变量,且因变量和自变量之间是线性关...原创 2019-02-03 11:39:29 · 8253 阅读 · 0 评论 -
利用DQN实现迷宫寻路
从q-learning到Deep-Q-Network笔者在利用q-learning算法在Tkinter模块中进行自主寻路,由于网格维度过大而导致训练结果不理想,设置1500次迭代再反复试验后发现q-learning算法在面临大量学习数据集问题上表现出结果不尽如人意,进而利用DQN的方法完成训练过程。再次特别鸣谢github博主(https://github.com/wangzhuanxing1...原创 2019-02-03 14:48:28 · 7464 阅读 · 1 评论 -
机器学习算法之----SVM
SVM代码实例本博客重在讲解SVM原理,以及其具体实现方式。What’s SVM?Support Vector Machine (SVM) is a supervised machine learning algorithm which can be used for both classification or regression. However, it is mostly used...原创 2019-02-04 01:03:42 · 974 阅读 · 0 评论 -
机器学习之----数据预处理
数据预处理方法实现Step 1: Importing the required LibrariesThese Two are essential libraries which we will import every time.NumPy is a Library which contains Mathematical functions.Pandas is the library u...原创 2019-02-06 21:04:34 · 420 阅读 · 0 评论 -
机器学习算法之----Decision Tree Classification
Decision Tree Classification决策树以树结构的形式构建分类或回归模型。它将数据集分解为越来越小的子集,同时逐步开发相关的决策树。最终结果是具有决策节点和叶节点的树。决策节点(例如,Outlook)具有两个或更多个分支(例如,Sunny,Overcast和Rainy)。叶节点(例如,Play)表示分类或决定。树中最顶层的决策节点,对应于称为根节点的最佳预测器。决策树可以处...原创 2019-02-06 21:30:45 · 1134 阅读 · 0 评论 -
机器学习算法之----Random Forests
Random Forests随机森林或随机决策森林是用于分类,回归和其他任务的集成学习方法,其通过在训练时构建多个决策树并输出作为类的模式(分类)或平均预测(回归)的类来操作。个别树木。随机决策森林纠正决策树过度拟合其训练集的习惯。随机决策森林的第一个算法是由Tin Kam Ho 使用随机子空间方法创建的,在Tin Kam Ho的公式中,这是一种实现Eugene Kleinberg提出的“随机...原创 2019-02-06 22:19:57 · 544 阅读 · 0 评论 -
自然语言处理1(NLP)------NLP--Basic Embedding Model
NLP–Basic Embedding Model1-1. NNLM(Neural Network Language Model) - Predict Next WordPaper下载代码块NNLM-Tensor.py@Tae Hwan Jung @graykodeimport tensorflow as tfimport numpy as nptf.reset_d...原创 2019-02-07 09:01:11 · 617 阅读 · 0 评论 -
自然语言处理2------CNN(Convolutional Neural Network)
CNN(Convolutional Neural Network)2-1. TextCNN - Binary Sentiment ClassificationPaper下载:TextCNN-Tensor.py''' code by Tae Hwan Jung(Jeff Jung) @graykode Reference : https://github.com/ioat...原创 2019-02-07 09:19:36 · 2428 阅读 · 0 评论 -
自然语言处理之----RNN(Recurrent Neural Network)
循环神经网络3-1. TextRNN - Predict Next StepPaper Finding Structure In TimeTextRNN-Tensor.py''' code by Tae Hwan Jung(Jeff Jung) @graykode'''import tensorflow as tfimport numpy as nptf.reset...原创 2019-02-07 09:33:33 · 563 阅读 · 1 评论 -
自然语言处理之------Attention Mechanism
Attention Mechanism4-1. Seq2Seq - Change WordPaper Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation(2014)''' code by Tae Hwan Jung(Jeff Jung) @gra...原创 2019-02-07 09:41:05 · 391 阅读 · 0 评论 -
自然语言处理之------ Model based on Transformer
Model based on TransformerDependencies:Python 3.5+Pytorch 0.4.1+5-1. The Transformer - TranslatePaper BERT: Pre-training of Deep Bidirectional Transformers for Language UnderstandingTran...原创 2019-02-07 09:50:46 · 947 阅读 · 0 评论 -
机器学习之------信号处理(入门原理)
TensorFlow Signal ProcessingProblem Overview该项目的目标是评估使用深度学习进行音频效果仿真的有效性。对于音频,有两种主要的工具分类:生成器和效果。生成器是接收非音频输入(物理或midi)并从中创建音频的东西。这将包括乐器,合成器,鼓 - 基本上任何真正突出传统音乐的东西。另一个类别,即效果,是将音频作为输入并将其转换为其他音频输出的元素。这可以是从...原创 2019-02-07 10:32:04 · 20176 阅读 · 0 评论 -
q-learning精讲
Q-learningQ-learning是一种用于机器学习的强化学习技术。 Q-learning的目标是学习一种策略,告诉Agent在什么情况下要采取什么行动。 它不需要环境模型,可以处理随机转换和奖励的问题,而无需进行调整。对于任何有限马尔可夫决策过程(FMDP),Q学习找到一种最优的策略,即从当前状态开始,它在任何和所有后续步骤中最大化总奖励的预期值。在给定无限探索时间和部分随机策略的情况...原创 2019-02-07 16:53:42 · 12351 阅读 · 0 评论 -
机器学习算法之----BP神经网络
BP神经网络1、神经网络model先介绍个三层的神经网络,如下图所示输入层(input layer)有三个units( X。为补上的bias,通常设为1)表示第j层的第i个激励,也称为为单元unit为第j层到第j+1层映射的权重矩阵,就是每条边的权重所以可以得到:隐含层:输出层其中,S型函数也成为激励函数可以看出为3x4的矩阵,为1x4的矩阵==》j+1的单元数x...原创 2019-02-09 19:53:25 · 786 阅读 · 0 评论 -
机器学习算法之------PCA主成分分析(降维)
PCA主成分分析(降维)1、用处数据压缩(Data Compression),使程序运行更快可视化数据,例如3D–>2D等2、2D–>1D,nD–>kD如下图所示,所有数据点可以投影到一条直线,是投影距离的平方和(投影误差)最小注意数据需要归一化处理思路是找1个向量u,所有数据投影到上面使投影距离最小那么nD–>kD就是找k个向量u(1),u(2)…u...原创 2019-02-09 23:00:28 · 694 阅读 · 0 评论 -
机器学习算法之------异常检测 Anomaly Detection
异常检测 Anomaly Detection1、高斯分布(正态分布)Gaussian distribution分布函数:其中,u为数据的均值,σ为数据的标准差σ越小,对应的图像越尖参数估计(parameter estimation)2、异常检测算法例子训练集:,其中假设相互独立,建立model模型:过程选择具有代表异常的feature:xi参数估计:计算p...原创 2019-02-09 23:18:25 · 1235 阅读 · 0 评论 -
机器学习之------监督、非监督、半监督
Unsupervised learning, Supervised learning and Semi-supervised learningIn the context of machine learning, what is the difference between unsupervised learning, supervised learning and semi-supervise...原创 2019-02-10 08:49:39 · 476 阅读 · 0 评论 -
自然语言处理之-----Word2Vec
A Beginner’s Guide to Word2Vec and Neural Word EmbeddingsIntroduction to Word2VecWord2vec是一个处理文本的双层神经网络。它的输入是一个文本语料库,它的输出是一组向量:该语料库中单词的特征向量。虽然Word2vec不是深度神经网络,但它将文本转换为深网可以理解的数字形式。 Deeplearning4j实现...原创 2019-02-22 22:33:06 · 1699 阅读 · 1 评论 -
机器学习算法之------LSTM
Understanding LSTM NetworksRecurrent Neural NetworksHumans don’t start their thinking from scratch every second. As you read this essay, you understand each word based on your understanding of pre...原创 2019-02-23 07:59:11 · 3950 阅读 · 0 评论 -
使用 Doc2Vec & Logistic Regretion 进行多类文本分类
The goal is to classify consumer finance complaints into 12 pre-defined classes using Doc2Vec and Logistic RegressionDoc2vec is an NLP tool for representing documents as a vector and is a generalizin...原创 2019-02-23 11:28:34 · 1337 阅读 · 0 评论 -
An intuitive introduction to Generative Adversarial Networks (GANs)
生成对抗网络(GAN)的直观介绍Warm upLet’s say there’s a very cool party going on in your neighborhood that you really want to go to. But, there is a problem. To get into the party you need a special ticket — t...原创 2019-02-23 18:33:16 · 365 阅读 · 0 评论 -
DCGANs (Deep Convolutional Generative Adversarial Networks)
Deep Convolutional Generative Adversarial NetworksGenerative Adversarial Networks最有趣的部分之一是Generator网络的设计。 发电机网络能够获取随机噪声并将其映射到图像中,使得鉴别器无法分辨哪些图像来自数据集以及哪些图像来自发生器。这是神经网络的一个非常有趣的应用。 通常,神经网络将输入映射到二进制输出(1或...翻译 2019-02-24 09:32:11 · 1530 阅读 · 0 评论 -
使用机器学习模型推荐航空公司承运人
Using Machine Learning Models to recommend Airline CarrierThere is no silver bullet when it comes to solving a problem using machine learning models. I am yet to come across a model that serves as o...原创 2019-02-24 19:14:46 · 342 阅读 · 0 评论 -
A Comprehensive Introduction to Different Types of Convolutions in Deep Learning
Towards intuitive understanding of convolutions through visualizationsIf you’ve heard of different kinds of convolutions in Deep Learning (e.g. 2D / 3D / 1x1 / Transposed / Dilated (Atrous) / Spatial...翻译 2019-02-24 20:52:09 · 1413 阅读 · 0 评论 -
使用Python对Anthem的游戏发布进行情感分析
Video game launches are plagued by drama. From misleading pre-order bundles, to games that are far from complete at launch, big publishers have quite a bit of risk to manage when it comes to deciding ...翻译 2019-02-24 23:06:31 · 813 阅读 · 0 评论 -
强化学习:从网格世界到自动驾驶
0. Agents, Environments, and Rewards在过去几年中,人工智能研究人员发表的许多主要公告都是一门称为强化学习(RL)的学科。 最近的突破主要是由经典RL创意的微小转折所驱动,这得益于强大的计算硬件和利用所述硬件的软件的可用性。为了了解现代深度RL模型对于计算的饥饿程度,下表是最近RL进展的非详尽集合以及完成每项任务所需的计算资源的估计。请注意,上表中的任务都经...原创 2019-02-25 11:00:32 · 1552 阅读 · 0 评论