data-science python data-preprocessing

Feature Scaling: Quick Introduction and Examples using Scikit-learn

27 Sep 2017   Feature Scaling techniques (rescaling, standardization, mean normalization, etc) are useful for all sorts of machine learning approaches and critical for things like k-NN, neural networks and anything that uses SGD (stochastic gradient descent), not to mention text processing systems.

Included examples: rescaling, standardization, scaling to unit length, using scikit-learn.

Read More ›

python data-visualization data-newsletter-5

Matplotlib, Pyplot, Pylab etc: What's the difference between these and when to use each?

26 Sep 2017   Do you often get confused with terms like maptlotlib, pyplot, pylab, figures, axes, gcf, gca, etc and wonder what they mean? Matplotlib is the toolkit, PyPlot is an interactive way to use Matplotlib and PyLab is the same thing as PyPlot but with some extra shortcuts.

Read More ›

data-newsletter-5 data-science best-practices

5 Tips for moving your Data Science Operation to the next Level

26 Sep 2017   Principles for disciplined data science include: Discoverability, Automation, Collaboration, Empowerment and Deployment.

Read More ›

recommender-systems data-newsletter-5

Highlights of the Talk with Dr. Konstan on Recommender Systems

24 Sep 2017   Some highlights of the Podcast Episode with Dr. Joseph Konstan on interesting topics related to Recommender Systems. Discussed topics include serendipity, serpentining, diversity and temporal effects.

Read More ›

python data-visualization plotting

Seaborn by Example: Data Visualization and Plotting using Python

09 Sep 2017   Seaborn is a higher-level interface to Matplotlib. It has a more convenient API and has useful data visualization functions right out of the box.

Read More ›

data-newsletter-5 data-science

Data Provenance: Quick Summary + Reasons Why

07 Sep 2017   Data Provenance (also called Data Lineage) is version control for data. It refers to keeping track of modifications to datasets you use and train models on. This is crucial in data science projects if you need to ensure data quality and reproducibility.

Read More ›

data-newsletter-4 recommender-systems

Lessons from the Netflix Prize: Changing Requirements and Cost-Effectiveness

04 Sep 2017   Netflix never really used the #1 winning solution to the Netflix Challenge. Some of the reasons were that just wasn't cost-effective to implement the full thing and another was that requirements had changed.

Read More ›

data-newsletter-4 kaggle data-science

Winning Solutions Overview: Kaggle Instacart Competition

04 Sep 2017   The Instacart "Market Basket Analysis" competition focused on predicting repeated orders based upon past behaviour. Among the best-ranking solutings, there were many approaches based on gradient boosting and feature engineering and one approach based on end-to-end neural networks.

Read More ›

technology   data-newsletter-4 machine-learning

A Quick Summary of Ensemble Learning Strategies

01 Sep 2017   Ensemble learning refers to mixing the outputs of several classifiers in various ways, so as to get a better result than each classifier individually.

Read More ›

technology   data-newsletter-4 machine-learning

Evaluation Metrics for Classification Problems: Quick Examples + References

31 Aug 2017   There are multiple ways to measure your model's performance in machine learning, depending upon what objectives you have in mind. Some of the most important are Accuracy, Precision, Recall, F1 and AUC.

Read More ›