Deep Learning with C#, .Net and Kelp.Net

The Ultimate Kelp.Net Deep Learning Guide

Matt R. Cole

• Editore: BPB Publications
• Anno: 2019
• ISBN:
  9789388511018
• Argomento: Informatica
• Categoria: Reti neurali

Leverage SharePoint Online Modern Experience to create beautiful, dynamic and mobile-ready sites and pages Description
Lots of small, medium and large organizations or enterprises are using Office 365 for their business. And Microsoft is also investing heavily on Office 365 and providing lots of new features in Office 365 and other services in Office 365 like Office application or SharePoint Online, Yammer, Teams, Flow or PowerApps, etc. SharePoint is one of the popular portal technologies and web-based business collaboration and document management system. With Office 365 subscription, organizations can use SharePoint Online. Microsoft has announced the Modern features in SharePoint for a long time. Modern Experience is the future of SharePoint Online and on-premises also. This book is a comprehensive guide that lets you explore the Modern features in SharePoint Online or SharePoint Server 2019. In the book, I have covered details on Modern Team sites, communication sites, how you can customize the team sites according to your business requirement. You will also get hands-on Experience on how you can customize Modern site pages. I have also explained in detail various new features of Modern list and document libraries in SharePoint. This book also contains a few SharePoint portal examples, you will get in-depth knowledge on how to design team sites with various useful web parts. Few Organizations are still using SharePoint On-premises versions like SharePoint server 2019. I have also explained the Modern Experience in SharePoint 2019. Always it is better to know also, what are the things which are not possible in SharePoint Modern Experience, based on which you can check the impact, before moving to the SharePoint Online Modern Experience. Audience
This book is for the site owners, power users or administrators who want to design attractive pages for SharePoint Modern team sites or publishing sites. Though the book is intended for SharePoint developer knowledge, but a little understanding of SharePoint is required. We have provided detailed steps with proper screenshots for references. This book is also for the developers who are trying to build pages for Modern SharePoint team sites or publishing site in SharePoint Online or SharePoint server 2019. What you will Learn
In this book, you will learn what are Modern Experiences in SharePoint. How we can handle at the organizational level. What are the things which are not possible in SharePoint Online Modern Experience. Various new features of SharePoint Online Modern list and document libraries.You will also learn various web parts and how we can use those web parts while designing pages for your sites. Various examples of SharePoint Modern portal designs. How we can create and customize Modern site pages. How we can also start with SharePoint Server 2019 and use various Modern web parts in SharePoint 2019 sites. Key Features

• Learn how to use SharePoint Online Modern Experience (Modern UI)
• Create a Modern team site and communication site for your organization in SharePoint Online or SharePoint Server 2019
• Effectively use Modern list and Libraries in SharePoint Online or SharePoint 2019
• Learn about various Modern SharePoint web parts
• Create attractive and responsive portals in SharePoint Online or SharePoint 2019

Table of Contents

• Data Science Fundamentals
• Installing Software and Setting up
• Lists and Dictionaries
• Function and Packages
• NumPy Foundation
• Pandas and Dataframe
• Interacting with Databases
• Thinking Statistically in Data Science
• How to import data in Python?
• Cleaning of imported data
• Data Visualization
• Data Pre-processing
• Supervised Machine Learning
• Unsupervised Machine Learning
• Handling Time-Series Data
• Time-Series Methods
• Case Study – 1
• Case Study – 2
• Case Study – 3
• Case Study – 4
• About the Author
  Bijaya is a Microsoft MVP (Office Servers & Services) and having more than 11 years of experience in Microsoft Technologies specialized in SharePoint. He is Co-founder of TSInfo Technologies, a SharePoint consulting, training & development company in Bangalore, India. He has been a technology writer for many years and writes many SharePoint articles on his websites SharePointSky.com and EnjoySharePoint.com. Bijaya is a passionate individual who loves public speaking, blogging and training others to use Microsoft products. Before co-founding TSInfo Technologies, he was working with small and large organizations in various SharePoint On-premises as well as SharePoint Online office 365 & various related technologies. Bijaya also likes to publish SharePoint videos on his EnjoySharePoint YouTube Channel.
  

CHAPTER 1

Take This ___
and ___ It

Even though we are meeting for the first time, I can guess the two words that you have in mind to fill in the blanks for this sentence. How can that be? And not only that, I bet that it is not just you and I that have picked the same two words. The chances are high that many of us have. How could it be that the chances of all of us picking the same two words without thinking about it are so very high? I am sure we can all recall hearing this phrase somewhere in the past (and maybe one or two of us said it…), but is the pattern recognition in our brains so good that these two words immediately were the ones that came to mind when we saw a partial implementation of that sentence? And how could so many of us, seemingly unconnected from each other, arrive at the same two words and complete the sentence in the same way in probably close to the same amount of time? Did we just witness a marvelous example of unsupervised deep learning at its finest or was it something else? How can such advanced pattern recognition ability happen so fast and be so predictable among a huge population of people? Or maybe it was reinforcement learning that brought this to the mind (Lord knows I have received negative reinforcement for saying it!). Regardless of how, something marvelous surely has just happened.

Let us think about this for a second, shall we? We have an area in our brains called the hippocampus and this area stores episodic memories (stuff that happened to you and me). Within the hippocampus, many neurons just fired when we saw that sentence, and the exciting neurons which they were connected to and so on, which made them fire until the pattern was complete. And all this happened in the blink of an eye. But did any of us have to give any thought into having to complete that sentence, or did it seem to pop up from our subconscious mind with the greatest of ease? The truth is that we may never know, at least not fully, but theories are abounded.

So how many of you work in the ‘real world’ as developers? For those who do, you know more often than not we walk through the doors, put our fireman’s hat on, and off we go. How we long for some pure R & D time where we can work on our code, create better algorithms, enhance the UI we have been putting off for what seems like forever. Ah! To be like the academic world for even the smallest amount of time! But alas, we are in the real world! We have products to deliver and timelines to meet. Some of us work around an agile approach to development, some do not. Everyone seems to be doing the same things, albeit a little bit differently.

With this in mind, let us start with a short story. I know there are some of you out there for whom this will hit home for sure. As we are embarking on a path down the deep learning lane, for those of you who are developers coming from the academic world, it is just a little bit different out here. But since we are talking about developing deep learning applications, it is time both the sides of the spectrum (developers and the academic world) meet! After all, we are about to start developing deep learning applications together and deliver them to our customers, so sit back and enjoy the early morning meeting. We will no doubt have many more like them!

Early one morning, I walked into my boss’s office for our daily morning meeting and could not help but notice that this time the room was full of people I had not seen before. No doubt! This was going to be a very important meeting I thought to myself, so I quickly took my seat, smiled and made the courtesy introductions while I awaited the start of the meeting. Today, I was happy because I was going to use Kelp.Net to start a new project. Our development team has little to no deep learning experience per se, neither on the academic nor the production side, so using Kelp. Net will allow us to make rapid progress in our project, something the boss would like to see. Until this point, he was a bit reluctant to embark on machine learning for our projects; he believed there was more hype than reality to a lot of it, and I could see his point.

Like most development shops, we do not write our production level code in R or Python; we are a Microsoft shop heavy on the C# .Net side, with some occasional Visual Basic and Java thrown in. Your chances of getting anything non-Microsoft related from our IT shop is somewhere between slim and none, which is no different from many places I am sure.

The boss began the meeting by saying, “We have been told from the upper management that we need to keep up with the times, so we will machine-learn our next project.” Nice, did not see that one coming. He continued with the introduction of the new data science team, which he told us has been put together to ensure that the company is headed in the right direction. We were also told how the company was100% behind this initiative of integrating AI, machine learning and deep learning technologies in our daily lives, starting with this project, and how all of us need to do our part to ensure its success. Then, the data science team began to speak.

The first words out of their mouths were R and Python. The boss slowly nodded his head in agreement. He thought he could not even get the latest version of Visual Studio from the support team without everyone on the planet needing to approve the request, and suddenly we are going to switch programming languages?

The data science team completed their presentation and the meeting disbanded for the day. Everyone looked at each other as they so often did, and we walked out of the office. My mind was now in an alternate universe with many questions swirling around in my head. I was thinking to myself:

1.	My boss wanted to use machine learning to solve problems, but of course, none of our projects have really good problem definitions.
2.	My boss, as well as the rest of the senior management, has been told how simple yet important deep learning is. We cannot let our competitors go ahead of us. “Just a combination of massive amount of data, cloud technologies and coding,” he said. How hard is it in my corporate world to request even the smallest amount of data from an external owner, yet now I need access to reams and reams of it across multiple owners? Hopefully, he and others will understand that to do this properly, we will need to spend a considerable amount of time locating, cleaning and preparing the data to be useful for this project, and that is before we even start coding! First, however, is getting the permissions from each data silo owner to access the data.
3.	Once I do get access to the 75 petabytes of data the company has been storing for 15 years, how in the world am I going to fit it into this wonderful, third hand-me-down laptop that IT support has given me? 2 GB RAM was the standard 10 years ago, but not really sure if the machine learning objective is going to work out so well like this. I will surely have a lot of time to start drinking coffee!
4.	I tried and let our data scientist team know why we might need to consider some alternative routes, but of course that explanation fell on deaf ears because they all have PhDs!
5.	Our models do and will fail while we are perfecting them, and that is an iterative part of how we make things work. This was going to be fun when confronted with why is it failing and trying to explain it to the upper management! I was hoping that the new data science team would be there to help explain to my boss why model failure is a part of the process, but they were insisting that we rewrite everything in R or Python because that is what real deep learning projects use.

If this sounds like a day in your development shop, in whole or in part, welcome to the crowd! This short story was meant to illustrate some of what happens when we take things from the academic world and become buzzword-compliant in the corporate world. Real-world software development and academia are many light years apart in things we encounter and how we must perform daily, especially when it comes to machine and deep learning. Most of us long for the pristine world of academia or even research and development. For most of us, we put the fireman’s hat on the minute we walk through the doors and off we go.

I am happy to present you two very powerful and flexible tools: Kelp.Net and ReflectInsight. I will show you a great framework that provides power and flexibility for C# .Net developers, while doing a great job of bridging the corporate and academic worlds. I will show you how to use the software, as well as give you a high-level overview of machine and deep learning to help you to be able to have meaningful conversations with that new team of data scientists. For those concerned as to whether the software may meet your needs, as Kelp.Net is open source, let me tell you that you and your team will have complete control over customizing it to handle any situation you may encounter.

Objectives of this book

In this book, we will discuss how you, the C# developer, can add amazing and powerful deep learning techniques to your applications. We will do so by presenting the powerful, flexible and extendable open source software package known as Kelp.Net. We will also present ReflectInsight, an incredibly powerful, flexible and robust logging framework which will prove to be invaluable when debugging your machine learning models and processes

INFO: Kelp.Net – Original Japanese version from Twitter: https://twitter.com/harujoh

English version by Matt R. Cole

ReflectInsight – ReflectSoftware

While ReflectInsight is not open source, the value that it adds to redefine real-time logging makes it an easy choice for any project where we need deep instrumentation and real-time logging.

But perhaps even more important than all of that, you will have an open source software package in Kelp.Net which you can enhance and embellish to create new and exciting versions with incredibly powerful features. You will also have an incredibly powerful logging framework which is of immense importance when it comes to debugging your models and algorithms. For both, I have created what I hope will be your de facto standard reference manuals to use and to help you understand what each tool has to offer individually as well as in tandem.

We will start with a quick chat about neural networks, machine and deep learning. We will talk more about some terms and concepts you need to be familiar with and some of the similarities and the differences between the two. Since our focus is on deep learning, we will place great emphasis on it. Next, we will talk about logging, the importance of logging, and what ReflectInsight can do for you. With all this in place, the stage will be set for us to deep dive into Kelp.Net and talk about its terminology, what it means, and how it can be used. In our last few sections, we will discuss about Kelp.Net deep learning models, how you can train and test them, and how you can write model tests yourself.

Let us be honest. Deep learning is a rage nowadays, isn’t it? It is the latest in a series of buzzwords being used across the corporate world and is on every billboard and advertisement across the land. We are flooded with targeted advertisements about machine learning and AI on our desktops every time we do a search for something. What used to be a pristine environment within academia has now escaped and been embraced by the corporate world. And what a difference that has made. Luckily for us, our increasing and...

Indice dei contenuti

• Cover
• Deep Learning with C#, .NET and Kelp.NET
• Copyright
• About the Author
• Reviewer
• Preface
• Acknowledgement
• Errata
• Table of Contents
• 1. Take This ___ and ___ It
• 2. Machine Learning/Deep Learning Terms and Concepts
• 3. Deep Instrumentation Using ReflectInsight
• 4. Kelp.Net Reference
• 5. Model Testing and Training
• 6. Loading and Saving Models
• 7. Sample Deep Learning Tests
• 8. Creating Your Own Deep Learning Tests
• Appendix A
• Appendix B
• OpenCL