随笔分类 - 调参(转载)
摘要:Neural Network HyperparametersMost machine learning algorithms involve “hyperparameters” which are variables set before actually optimizing the model'...
阅读全文
摘要:11 Clever Methods of Overfitting and how to avoid themOverfitting is the bane of Data Science in the age of Big Data. John Langford reviews "clever" m...
阅读全文
摘要:Deep Learning in a Nutshell: History and TrainingThis series of blog posts aims to provide an intuitive and gentle introduction to deep learning that ...
阅读全文
摘要:FITTING A MODEL VIA CLOSED-FORM EQUATIONS VS. GRADIENT DESCENT VS STOCHASTIC GRADIENT DESCENT VS MINI-BATCH LEARNING. WHAT IS THE DIFFERENCE?In order ...
阅读全文
摘要:How much training data do you need?//@樵夫上校: 0. 经验上,10X规则(训练数据是模型参数量的10倍)适用与大多数模型,包括shallow network. 1.线性模型可以应用10X的经验规则,模型参数是特征选择后的数量(PCA等方法)。2.NN可以将10...
阅读全文
摘要:Recruit Coupon Purchase Winner's Interview: 2nd place, Halla YangRecruitPonpareis Japan's leading joint coupon site, offering huge discounts on everyt...
阅读全文
摘要:CIKM Competition数据挖掘竞赛夺冠算法陈运文背景CIKM Cup(或者称为CIKM Competition)是ACM CIKM举办的国际数据挖掘竞赛的名称。CIKM全称是International Conference on Information and Knowledge Mana...
阅读全文
摘要:Training (deep) Neural Networks Part: 1Nowadays training deep learning models have become extremely easy with high-quality libraries such as Torch and...
阅读全文
摘要:Liberty Mutual Property Inspection, Winner's Interview: Qingchen WangThe hugely popular Liberty Mutual Group: Property Inspection Predictioncompetitio...
阅读全文
摘要:ICDM Winner's Interview: 3rd place, Roberto DiazThis summer, theICDM 2015conferencesponsoredacompetitionfocusedon makingindividual user connections ac...
阅读全文
摘要:Diabetic Retinopathy Winner's Interview: 1st place, Ben GrahamBen Graham finished at the top of the leaderboard in the high-profileDiabetic Retinopath...
阅读全文
摘要:Training Neural Networks: Q&A with Ian Goodfellow, GoogleNeural networks require considerable time and computational firepower to train. Previously, r...
阅读全文
摘要:Machine Learning Trick of theDay (2): Gaussian Integral TrickToday's trick, theGaussian integral trick, is one that allows us to re-expressa (potentia...
阅读全文
摘要:Machine Learning Trick of theDay (1): Replica Trick'Tricks' of all sorts are used throughoutmachine learning, in both research and in production setti...
阅读全文
摘要:Bayesian optimisation for smart hyperparameter searchFitting a single classifier does not take long, fitting hundreds takes a while. To find the best ...
阅读全文
摘要:CIFAR-10 Competition Winners: Interviews with Dr. Ben Graham, Phil Culliton, & Zygmunt ZającDr. Ben GrahamDr. Ben Grahamis an Assistant Professor in S...
阅读全文
摘要:Deep Learning:Assuming a deep neural network is properly regulated, can adding more layers actually make the performance degrade?I found this to be re...
阅读全文
摘要:Cross Validation done wrongCross validation is an essential tool in statistical learning1to estimate the accuracy of your algorithm. Despite its great...
阅读全文
摘要:CrowdFlower Winner's Interview: 1st place, Chenglong ChenTheCrowdflower Search Results Relevancecompetition asked Kagglers toevaluate the accuracy of ...
阅读全文
摘要:Taxi Trip Time Winners' Interview: 3rd place, BlueTaxiThis spring, Kaggle hostedtwo competitions with theECML PKDD conferencein Porto, Portugal. The c...
阅读全文