tarunaditya Goto Github PK

alexnet-experiments-keras

Code examples for training AlexNet using Keras and Theano

convnets-keras

pretrained alexnets in keras

glove

GloVe model for distributed word representation

handling-class-imbalance

Often in classification problems have class imbalance eg problem of finding out propensity to click (CTR for marketing campaigns dont go beyond 9%) / Impressions viewed in a given spot. The typical classifiers when trained on such a data simply classifies a new training data point to either click or no click. In other words they would be having poor Precision rates/ specificity. There are a few ways to have a work around, one way is to introduce more hypothetical datapoints of the infrequent class, second way is to sample from more frequent class 'smartly', thirdly look at the code & its comments to understand more about certain new packages in R that i found interesting that does much better job with respect to AUC metrics

keras-spp

Spatial pyramid pooling layers for keras

location_entropy

generic module that generates location entropy based on location visits distributed across time & can be modified to do the same for visits distributed across users if the input data is available in that format

loldja.github.io

Data science blog

pythondatasciencehandbook

Python Data Science Handbook: full text in Jupyter Notebooks

redshift

R library that helps you connect to redshift tables

scala-machine-learning-projects

Scala Machine Learning Projects, published by Packt

spark_with-java-api-to-test-performance

streaming_with_scala_for_better_performance

scala performs well with respect to throughput on our infra, has better resource utilization like executor jvm sizes etc

survival_analysis

kaplan meier & lifelines based

timeseries

explores almost all packages/ tools that are used in timeseries modelling, ive tried to use various models depending on how well they perform, ive tried to explain logic (through the comment section) of choice of models based on metrics like acf, pacf etc.

twittercorpus_wordembeddings_basemodel_word2vec_generic_nlp_pipeline

creating wordembedding matrix on twitter data