Showing posts with label MachineLearning. Show all posts
Showing posts with label MachineLearning. Show all posts

Sunday, 26 June 2016

Book Review: Microsoft Azure Machine Learning by @sumitmund

Microsoft Azure Machine Learning by Sumit Mund is a good beginner level book that made Azure ML a little less daunting for me. I'll keep going back to it as I continue to learn & experiment more with Azure ML, the amazing predictive analysis tool that makes it easy for developers to leverage complex Machine Learning algorithms all within the web browser.

Having quickly read through the book to familiarize myself with Machine Learning jargon & tried some of the samples, I've realized it'll take me more reading and practice to achieve proficiency. This 200+ paged book whets your learning appetite by covering a breadth of topics with just enough information to get you started but not too much to make it too intimidating for newbies.

It helps that the author makes practical tips & recommendations rather than just pontificating. Sample -
Outliers are data points that are distinctly separate from the rest of the data... In many cases, it is a good idea to clip or remove the outliers.

As with anything related to the Cloud, most info has a limited shelf-life - a serious threat to books & documentation. As Azure ML Studio continues to evolve, some things have changed since the book was written. Therefore you'll find that some Modules described in the book have new names and it is an exercise for the reader to identify the differences.

A little more effort could have been spent in presenting the content better. While the author provides background context for most keywords before introducing them, there were a few instances where the author drops keywords without adequate explanation like in the case of "ports".

Despite the minor blemishes, this book is engaging and well-structured. It helped me appreciate the understated power of Azure ML.

Sunday, 3 April 2016

This Week I Learned - Week #156

This Week I Learned -

* Google Search uses Base64-encoded images. Base64 encoding will produce large output. The trade off with using Base64-encoded images in a web page is that you get instant image loading in exchange for inflating your page size slightly. What you don't want to be doing is using this for massive detailed images. Where it does shine is if your Web server is compressing output before delivering it. Because of the nature of Base64, it compresses exceptionally efficiently, so a dense amount of data like that will squash down to a very small size.

*  With HTTPS the hostname (api.somewebsite.com) is passed in the clear but the full URL is not (so it will not be susceptible to casual traffic sniffing attacks). In HTTPS, the actual URL is sent only after the SSL handshake has been performed. HTTPS uses TLS, which is Transport Layer Security. HTTP as a protocol, runs above the transport layer.

* A URL like "https://api.somewebsite.com/v1/something&key=YOUR-API-KEY" is commonly known as capability / secret URL. Capability URLs are a lot more secure than people give them credit for.

* Documents newly shared on Google Drive with links to third-party HTTPS websites, will no longer relay the original document’s URL

* In the age of Machine Learning it's important to treat all data as secure. It's just too easy to reveal minor details (like sleep schedules) from API emissions and other items traditionally thought of as "low risk."

* Before J K Rowling 's Harry Potter series was a commercial and critical hit, Rowling was struggling to make Potter's story come to fruition. Her manuscript was rejected by 12 different publishing houses before she finally found a publisher.

* Amazon Asia and Catamaran Ventures have jointly formed an entity called Cloudtail India Pvt. Ltd to sell items such as books, phones and exclusive Amazon merchandise on Amazon.in under the seller name Cloudtail, according to documents available with the Registrar of Companies (RoC). Cloudtail India is a fully owned unit of Prione Business Services Pvt. Ltd, the documents show. Prione, in turn, is owned by Catamaran Management Services Pvt. Ltd, the acting trustee of Hober Mallow Trust (51%), Amazon Asia Pacific Resources Pvt. Ltd (48%) and Amazon Eurasia Holdings (1%). Catamaran is the family-owned fund of Infosys Ltd co-founder Murthy. Amazon ..,Flipkart have...created a complex structure to try and comply with laws. WS Retail, a seller on Flipkart, was owned by Flipkart co-founders Sachin Bansal and Binny Bansal (unrelated) until September 2012 - Live Mint

* Since the start of 2016, oil prices have swung between $27 and $42 per barrel, about a quarter of the 2008 peak crude oil price of $145.  India’s fiscal deficit has improved since the country saved nearly $70 billion on importing crude and other petroleum products in 2015.