Skip to content

Data Science & Machine Learning Newsletter # 79

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group Pre­pro­cess­ing for Machine Learn­ing with tf.Transform When apply­ing machine learn­ing to real world datasets, a lot of effort is required to pre­process data into a for­mat suit­able for stan­dard machine learn­ing mod­els, .…  Today we are announc­ing tf.Transform, a library for Ten­sor­Flow that allows users to define pre­pro­cess­ing pipelines and run these using large scale data pro­cess­ing Devamını Oku […]

Data Science & Machine Learning Newsletter # 78

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group Machine Learn­ing in Cyber Secu­rity Domain 1: Fun­da­men­tals 2: Cyber Secu­rity Rat­ing and Inci­dent Fore­cast­ing 3: Fraud Detec­tion 4: Secure User Authen­ti­ca­tion NanoNets : How to use Deep Learn­ing when you have Lim­ited Data  One com­mon bar­rier for using deep learn­ing to solve prob­lems is the amount of data needed to train a model. Anonymiz­ing doc­u­ments with Word Vec­tors Devamını Oku […]

Data Science & Machine Learning Newsletter # 77

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group Train­ing a deep learn­ing model to steer a car in 99 lines of code Deep learn­ing in 2017 is mag­i­cal. We get to apply immensely com­plex algo­rithms to equally com­plex prob­lems with­out hav­ing to spend all our time writ­ing the algo­rithms our­selves. Instead, thanks to libraries like Ten­sor­Flow and Keras, we get to focus on the fun stuff: model archi­tec­ture, para­me­ter tun­ing and data aug­men­ta­tion. Com­pre­hen­sive Devamını Oku […]

Data Science & Machine Learning Newsletter # 76

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group Domain Trans­fer Net­work (DTN) Ten­sor­Flow imple­men­ta­tion of Unsu­per­vised Cross-Domain Image Gen­er­a­tion. Mas­ter­ing the Game of Go with Deep Neural Net­works and Tree Search Google Deep­Mind paper A Sur­vey on Trans­fer Learn­ing A major assump­tion in many machine learn­ing and data min­ing algo­rithms is that the train­ing and future data must be in the same fea­ture space and Devamını Oku […]

Data Science & Machine Learning Newsletter # 75

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group Pre­dict­ing with con­fi­dence: the best machine learn­ing idea you never heard of One of the dis­ad­van­tages of machine learn­ing as a dis­ci­pline is the lack of rea­son­able con­fi­dence inter­vals on a given pre­dic­tion. There are all kinds of rea­sons you might want such a thing, … , for exam­ple, that some­one will click on an ad, you prob­a­bly want to serve one that pays a nice click through rate. Devamını Oku […]

Data Science & Machine Learning Newsletter # 74

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group https://github.com/100/Cranium/ A portable, header-only, arti­fi­cial neural net­work library writ­ten in C99 king — man + woman is queen; but why? word2vec is an algo­rithm that trans­forms words into vec­tors, so that words with sim­i­lar mean­ing end up lay­ing close to each other. More­over, it allows us to use vec­tor arith­metics to work with analo­gies, for exam­ple the famous king Devamını Oku […]

Data Science & Machine Learning Newsletter # 73

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group 4 Rea­sons Your Machine Learn­ing Model is Wrong (and How to Fix It) In this post we’ll walk through some com­mon sce­nar­ios where a seem­ingly good machine learn­ing model may still be wrong. We’ll show how you can eval­u­ate these issues by assess­ing met­rics of bias vs. vari­ance and pre­ci­sion vs. recall, and present some solu­tions that can help when you encounter such sce­nar­ios. How Devamını Oku […]

Data Science & Machine Learning Newsletter # 72

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group Best Data Visu­al­iza­tion Projects of 2016 s of 2016 Big Pic­ture: Google Visu­al­iza­tion Research The Unrea­son­able Effec­tive­ness of Recur­rent Neural Net­works We’ll train RNNs to gen­er­ate text char­ac­ter by char­ac­ter and pon­der the ques­tion “how is that even pos­si­ble?” Python Scikit-learn to sim­plify Machine learn­ing : { Bag of words } To [ TF-IDF ] Scikit-learn Devamını Oku […]

Data Science & Machine Learning Newsletter # 71

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group im2latex This is a Ten­sor­Flow port of harvardnlp’s im2markup project. The model accepts images of math­e­mat­i­cal equa­tions type­set in Latex, and out­puts the markup used to gen­er­ate those images. 50 things I learned at NIPS 2016 Quick notes on NIPS 2016 Con­fer­ence Deep Learn­ing for NLP https://github.com/airbnb/superset Super­set is a data explo­ration plat­form designed Devamını Oku […]

Data Science & Machine Learning Newsletter # 70

You want to get updates? Please join Data Sci­ence & Machine Learn­ing Newslet­ter Linked Group Character-level Con­vo­lu­tional Net­works for Text Clas­si­fi­ca­tion This arti­cle offers an empir­i­cal explo­ration on the use of character-level con­vo­lu­tional net­works (Con­vNets) for text clas­si­fi­ca­tion. Intro­duc­tion to Word­Net https://github.com/mbernico/snape/ Snape is a con­ve­nient arti­fi­cial dataset gen­er­a­tor that wraps sklearn’s make_classification and make_regression and then Devamını Oku […]