Machine-Learning

Now, why do this? What’s wrong with Medium and Substack and Blogger and Wordpress and all the tens of websites out there? Nothing. You could totally just create a blog there and publish. That’s why Content Management Systems exist - so you could focus on the content and not on the management . But if you’re like me and want some more control over how you publish your blogs, customize them etc. , you will find this and the other upcoming articles in this category interesting. ...

(Here’s the preface in case you missed it - [[Reading DDIA - Preface]]) Chapter 1 - Reliable, Scalable and Maintainable Applications This chapter’s goal is to effectively give us a high level overview of what this book is fundamentally about - Thinking about data systems along the axes of : Reliability Scalability Maintainability What is a reliable system? My thoughts before reading the section : A system that is available to return results/outputs that are expected of it when it is employed. If I’m accessing an API, if the API is available 99 out of 100 times, then its reliability is 99% If a server is available for 23 hrs 45 minutes per day then its availability is 23.75/24 = 98.9% But is availability == reliability? Let’s see. A reliable human is someone who does something that he says he will do or is expected to do every time. Similarly a reliable system is one that returns the expected results every time that they are used. An ATM is reliable if I can withdraw money from it in the denominations that I want, when I want it. For example, the ATM in Puttenahalli , JP Nagar is perpetually unavailable. Everytime I go there, the system is under maintenance. It is 0% reliable. Let’s read the chapter. ...

These are the commands I use a lot using Pandas - Get rows with NaNs in any column df[df.isna().any(axis=1)] Get dataframe excluding all NaNs in a particular Column df[df['<col_name>'].notnull()] Get counts of number of Nulls per column df.isnull().sum() Increase width of columns displayed in Jupyter Notebooks - This is possibly one of my most used commands for every notebook that involves dataframes. pd.set_option('display.max_colwidth',400)

I attended an interesting Webinar titled “A Data Science Playbook for Explainable ML/AI” conducted by Chief Data Scientist,Josh Poduska, and VP of Marketing, Jon Rooney of Domino Data. This post lays out some of the highlights (in my opinion) of the talk, peppered with my understandings and some additional points that might be of interest. I will breaking this learning-post up into two parts - the first part discussing the theoretical concepts of interpretability and Part-2 on different Open Source models that are in practice today. This is Part-1. ...