去年底，久违的丹·布朗大叔终于带着新作《Origin》回归了，距离上一本小说《地狱》隔了四年！而距离红遍大江南北的《达芬奇密码》，已经过去了近15年！！！！！！
小编的最爱是《天使与魔鬼》，中文版看过两遍以后，英文版和电影又各刷了一遍。话说刷英文版那段时间也经历着失眠的困扰，结果刷着刷着，养成了刷15分钟秒睡的好习惯...... 在此跑题推荐有失眠困扰的孩子们，都可以尝试一下丹布朗的安眠药。
作为忠实书迷的小编，当然是第一时间，买了本有声书，18个小时9分钟的时长。正好那段时间有些失眠，每天睡前开启15分钟有声书睡眠模式，即使小说情节越来越精彩，也抵挡不住我洪荒的睡意...... 就酱紫，每天睡前15分钟加一趟去加拿大的长途开车，把这部大部头听完了......
还是熟悉的丹·布朗风格，融合了密码学、科技、宗教、历史、艺术、建筑等元素...... 最让人眼前一亮的是，丹叔竟然抱起了人工智能的大腿。丹叔表示，近期一直在密切关注人工智能技术方面的进展。丹叔对人工智能的兴趣在新书中得到了很好的呈现，书中人工智能Winston简直聪明机智到令人发指......
还是熟悉的罗伯特·兰登配方，最大的不同大概是，老当益壮but风流倜傥、几乎每部都要换个兰登女郎的哈佛教授，这部没和女主发展爱情线了，毕竟女主是未来西班牙王后......
(汤姆汉克斯饰演了电影《天使与魔鬼》、《达芬奇密码》、《但丁密码》中的罗伯特·兰登)
据说，丹叔还会将兰登教授的故事继续写下去，而且下一部的探险目的地有可能是中国。毕竟在新书中，丹叔可是出乎意料地小舔了一下wuli清华大学。书中一名西班牙王室的公关，就在清华大学留过学。
《Origin》在亚马逊上的评分是4星，评价正负参半。总体来说，小编觉得这部作品值得3.8-4分的评价，想练英语听力和治疗失眠的，可以买有声书来听，科科。中文版据说将在今年春天上架，大家拭目以待！
话说回来，书迷们都在翘首期盼新作的这四年里，丹叔到底在干嘛呢？也许，去学人工智能了吧？
DAL将在2月24日
推出全新课程
《AI人工智能训练营》！
适合学员背景
理工科或者计算机Computer Science专业, 数学统计专业, 计算机编程爱好者
如果Python背景比较弱, 可以先参加我们的Python基础入门课程
三大模块:
机器学习, 深度学习与神经网络, 案例项目实践
双语教学:
机器学习Machine Learning部分全程由英文教学, 方便学员未来求职时对答如流. 深度学习与神经网络和项目实践部分才用中英双语
三大名师:
Peter(USC Information Institute Post Doc, MachineLearning), Carol (Google工程师, 精通Tensorflow), Eric(Google工程师, AI专家)
前沿技术:
课程设计多项AI领域必备前沿技术, 包括全面系统的Machine Learning知识讲解梳理, 神经网络与深度学习从入门到实践,Tensorflow实战入门, 人脸识别项目, NLP自然语言处理实战项目
实战演练:
课程内容基于实战项目, 边学习边练习 项目一: Facial Recogniztion 项目二:Natural Language Processing (详见syllabus)
-
全面系统了解Machine Learning(Regression, Classification,Dimension Reduction, Clustering)
-
了解Neutral Network与DeepLearning (Neural Network, deep neutral network, convolution neural network, 调参技巧, RNN等)
-
用Tensorflow实战FacialRecognition, 并且尝试改进与调参
-
学习和掌握深度学习与NLP自然语言处理 (NLP概念与基础, word2vec, GloVe, 复杂NLP模型)
-
用Tensorflow实战NLP项目
Modular 1 – Machine Learning
Class 1 Regression
-
Basic concept of Regression
-
Bias-Variance tradeoff
-
Underfitting vs.Overfitting
-
Linear regressionanalytical solution
-
Regularization: Lasso，Ridge，Elastic-Net，Pros and cons of L1and L2 regularization
-
Advanced techniquesin regression，Gradient Descendent，CoordinatedDescendent，Stochastic GradientDescendent，Random sampleconsensus (RANSAC)
Class 2 Classification
-
Evaluation Methodsof classification
-
Basicclassification model: logistic regression, decision tree
-
ClassificationTypes (how binary and multi-class works)
-
Ensemble modelmethod: Bagging，Boosting，Stacking
Class 3 DimensionReduction
-
Dimension reduction overview
-
Dimension reduction methods：Randomized Projection，Principal Component Analysis，PCA Calculation，Randomized PCA，Sparse PCA
-
Manifold learning
-
MultidimensionalScaling：MDS，Isomap
Class 4 Clustering
-
Unsupervisedlearning introduction
-
Clustering methods& techniques：K-mean Algorithm，HierarchicalClustering Algorithm，DBSCAN algorithm，
Outlierand anomaly detection
Modular 2 – NeuralNetwork & Deep Learning
Class 1
Neuralnetwork basic (which maybe duplicate with NLP part)
-
Introduction toneural network, include some basic concept like neuron, weights, bias,activation function.
-
Forward propagation for inference
-
Training algorithm: backpropagation (use 1hidden layer neural network with binary output as example)
Deep neural network
Convolutionalneural network
-
Motivation: why useCNN in computer vision problem: position invariance.
-
Convolution
-
Intro toconvolutional layer + pooling layer
-
Revisit MNIST problem and show how to use CNNto improve it. #params reduced.
Class 2 Short recap to fully connected layer, convolutional layer and pooling layer
.
Introduction tofamous vision problems and corresponding networks
Useful technicalsfor neural network training
-
Performance: trydifferent network structure, different number of layers and different number ofhidden units in each layer
-
Converge: sensitive to learning rate
-
Speed up training:Stochastic Gradient Descent, Momentum
-
Gradient vanishingproblem: Batch Normalization
Recurrent NeuralNetwork for video learning
Reinforcementlearning
Class 3 –Tensorflow and Facial Recognition
Brief introductionto Tensorflow: Tensor, operator concept.
Shows one smallnetwork structure and shows how to write it in Tensorflow.
Lab problem:
Face recognition:given face images for 40 person, each have 10 images, use 9 images of eachperson for training. Target is to label the left 40 images (1 per person) tothe right person.
Face recognition iswidely used technologies, such as photo softwares, surveillance.
Class 4
Finish the basic versionfor the Face recognition.
Improve network:
Modular 3 – DeepLearning and Natural Language Processing
Class 1
Class 2
-
Simple Word Vectorrepresentations: word2vec, GloVe
-
Vector(discrete)Representation
-
Problem withdiscrete representation
-
Cooccurence Matrix
-
Main idea ofword2vec
-
Main idea of Glove
