Our Technology Lead Cory Zibell gave a presentation about Machine Learning. The algorithms, processes, techniques, and modules that it entails. It's meant for anyone to grasp, check it out!
Innovations in technology has revolutionized financial services to an extent that large financial institutions like Goldman Sachs are claiming to be technology companies! It is no secret that technological innovations like Data science and AI are changing fundamentally how financial products are created, tested and delivered. While it is exciting to learn about technologies themselves, there is very little guidance available to companies and financial professionals should retool and gear themselves towards the upcoming revolution.
In this master class, we will discuss key innovations in Data Science and AI and connect applications of these novel fields in forecasting and optimization. Through case studies and examples, we will demonstrate why now is the time you should invest to learn about the topics that will reshape the financial services industry of the future!
Topic
- Frontier topics in Optimization
Stochastic computer models, wherein reruns of the code with the exact same inputs does not yield the exact same output, are becoming increasingly commonplace. Effective statistical analysis of such output can be more challenging and more crucial than the statistical analysis of deterministic computer models. Even so, stochastic simulation is currently subject to less statistical research focus.
This talk will outline a review we have been working on, in which we aim to spur additional research on the topic – introducing the objectives; outlining what statistical models currently exist; discussing how one can efficiently use such models to answer key questions about a stochastic computer model, and explaining what challenges currently still exist.
Leveraging Machine Learning or IA in order to detect Credit Card Fraud and suspicious transations. The aim of this presentation is to help you to improve your knowledge in Machnie Learning and to start development of multiple families of algorithms in Python.
Innovations in technology has revolutionized financial services to an extent that large financial institutions like Goldman Sachs are claiming to be technology companies! It is no secret that technological innovations like Data science and AI are changing fundamentally how financial products are created, tested and delivered. While it is exciting to learn about technologies themselves, there is very little guidance available to companies and financial professionals should retool and gear themselves towards the upcoming revolution.
In this master class, we will discuss key innovations in Data Science and AI and connect applications of these novel fields in forecasting and optimization. Through case studies and examples, we will demonstrate why now is the time you should invest to learn about the topics that will reshape the financial services industry of the future!
Topic
- Frontier topics in Optimization
Stochastic computer models, wherein reruns of the code with the exact same inputs does not yield the exact same output, are becoming increasingly commonplace. Effective statistical analysis of such output can be more challenging and more crucial than the statistical analysis of deterministic computer models. Even so, stochastic simulation is currently subject to less statistical research focus.
This talk will outline a review we have been working on, in which we aim to spur additional research on the topic – introducing the objectives; outlining what statistical models currently exist; discussing how one can efficiently use such models to answer key questions about a stochastic computer model, and explaining what challenges currently still exist.
Leveraging Machine Learning or IA in order to detect Credit Card Fraud and suspicious transations. The aim of this presentation is to help you to improve your knowledge in Machnie Learning and to start development of multiple families of algorithms in Python.
Building machine learning systems remains something of an art, from gathering and transforming the right data to selecting and finetuning the most fitting modeling techniques. If we want to make machine learning more accessible and foster skilfull use, we need novel ways to share and reuse findings, and streamline online collaboration. OpenML is an open science platform for machine learning, allowing anyone to easily share data sets, code, and experiments, and collaborate with people all over the world to build better models. It shows, for any known data set, which are the best models, who built them, and how to reproduce and reuse them in different ways. It is readily integrated into several machine learning environments, so that you can share results with the touch of a button or a line of code. As such, it enables large-scale, real-time collaboration, allowing anyone to explore, build on, and contribute to the combined knowledge of the field. Ultimately, this provides a wealth of information for a novel, data-driven approach to machine learning, where we learn from millions of previous experiments to either assist people while analyzing data (e.g., which modeling techniques will likely work well and why), or automate the process altogether.
AI and ML Skills for the Testing World TutorialTariq King
Software continues to revolutionize the world, impacting nearly every aspect of our work, family, and personal life. Artificial intelligence (AI) and machine learning (ML) are playing key roles in this revolution through improvements in search results, recommendations, forecasts, and other predictions. AI and ML technologies are being used in platforms for digital assistants, home entertainment, medical diagnosis, customer support, and autonomous vehicles. Testing practitioners are recognizing the potential for advances in AI and ML to be leveraged for automated testing—an area that still requires significant manual effort. Tariq King and Jason Arbon introduce you to the world of AI for software testing. Learn the fundamentals behind autonomous and intelligent agents, ML approaches including Bayesian networks, decision tree learning, neural networks, and reinforcement learning. Discover how to apply these techniques to common testing tasks such as identifying testable features, generating test flows, and detecting erroneous states.
Improvement of strip thickness control through the process of data analyticsSri Raghavan
o The aim of this research study is to perform data mining for the improvement of strip thickness control on a cold reduction mill using data analytics. This project is done for Cogent Power (a subsidiary company of Tata Steel) located at Newport, UK. Further to this, a software was developed in python to perform data mining to avoid developing codes for future purposes
Machine Learning presentation. Helps you to have a brief idea about what machine learning is and gives you direction to go deep into it. It covers the idea of Supervised learning and unsupervised learning and examples of how to use different models.
San Francisco Hacker News - Machine Learning for HackersAdam Gibson
This was for the san francisco hacker news meetup in february at engineyard.
This was intended as a basic intro to machine learning for people who wanted to step in to the field.
Video coming shortly.
Data Analytics, Machine Learning, and HPC in Today’s Changing Application Env...Intel® Software
This session explains how solutions desired by such IT/Internet/Silicon Valley etc companies can look like, how they may differ from the more “classical” consumers of machine learning and analytics, and the arising challenges that current and future HPC development may have to cope with.
Learning Graphs Representations Using Recurrent Graph Convolution Networks Fo...Yam Peleg
Recently, number of papers re-visited this problem of generalizing neural networks to work on arbitrarily structured graphs, some achieving promising results in domains that have previously been dominated by other shallower algorithms. While Graph convolutions are generalization of spacial convolutions, and easiest to define in spectral domain, General Fourier transform used to represent them scales poorly with size of data. Therefore, first order approximation in Fourier-domain used to obtain efficient linear-time graph-CNNs. Those scales poorly with size of data. due to that, the expressiveness power of the proposed graph convolutional networks is severely impoverished. Another approach for learning graph representations requires the repeated application of contraction maps as propagation functions until node representations reach a stable fixed point. We combine those approaches and propose a recurrent version of Relational Graph Convolution networks, we then proceed to construct two models, Recurrent Variational Graph AutoEncoder and Recurrent Graph Convolution Regressor and show that for Ethereum Blockchain transaction graph we outperform the traditional Graph Convolution Network at predicting future movments of the corresponding tradable asset: Ether.
Machine learning for sensor Data AnalyticsMATLABISRAEL
במצגת זאת נראה כיצד עושים Machine Learning בסביבת MATLAB. נציג מספר יכולות ואפליקציות מובנות ההופכות את תהליך למידת המכונה ליעיל ומהיר יותר – כלים כמו ה-Classification Learner, ה-Regression Learner ו-Bayesian Optimization. בהסתמך על מידע המתקבל מחיישני סמארטפון, נבנה מערכת סיווג המזהה את הפעילות שמבצע המשתמש – הליכה, טיפוס במדרגות, שכיבה, וכו'
Our guide will provide you with a roadmap of the current situation, what this means for brands, and what you can do in the coming months to protect your brand’s vitality.
Building machine learning systems remains something of an art, from gathering and transforming the right data to selecting and finetuning the most fitting modeling techniques. If we want to make machine learning more accessible and foster skilfull use, we need novel ways to share and reuse findings, and streamline online collaboration. OpenML is an open science platform for machine learning, allowing anyone to easily share data sets, code, and experiments, and collaborate with people all over the world to build better models. It shows, for any known data set, which are the best models, who built them, and how to reproduce and reuse them in different ways. It is readily integrated into several machine learning environments, so that you can share results with the touch of a button or a line of code. As such, it enables large-scale, real-time collaboration, allowing anyone to explore, build on, and contribute to the combined knowledge of the field. Ultimately, this provides a wealth of information for a novel, data-driven approach to machine learning, where we learn from millions of previous experiments to either assist people while analyzing data (e.g., which modeling techniques will likely work well and why), or automate the process altogether.
AI and ML Skills for the Testing World TutorialTariq King
Software continues to revolutionize the world, impacting nearly every aspect of our work, family, and personal life. Artificial intelligence (AI) and machine learning (ML) are playing key roles in this revolution through improvements in search results, recommendations, forecasts, and other predictions. AI and ML technologies are being used in platforms for digital assistants, home entertainment, medical diagnosis, customer support, and autonomous vehicles. Testing practitioners are recognizing the potential for advances in AI and ML to be leveraged for automated testing—an area that still requires significant manual effort. Tariq King and Jason Arbon introduce you to the world of AI for software testing. Learn the fundamentals behind autonomous and intelligent agents, ML approaches including Bayesian networks, decision tree learning, neural networks, and reinforcement learning. Discover how to apply these techniques to common testing tasks such as identifying testable features, generating test flows, and detecting erroneous states.
Improvement of strip thickness control through the process of data analyticsSri Raghavan
o The aim of this research study is to perform data mining for the improvement of strip thickness control on a cold reduction mill using data analytics. This project is done for Cogent Power (a subsidiary company of Tata Steel) located at Newport, UK. Further to this, a software was developed in python to perform data mining to avoid developing codes for future purposes
Machine Learning presentation. Helps you to have a brief idea about what machine learning is and gives you direction to go deep into it. It covers the idea of Supervised learning and unsupervised learning and examples of how to use different models.
San Francisco Hacker News - Machine Learning for HackersAdam Gibson
This was for the san francisco hacker news meetup in february at engineyard.
This was intended as a basic intro to machine learning for people who wanted to step in to the field.
Video coming shortly.
Data Analytics, Machine Learning, and HPC in Today’s Changing Application Env...Intel® Software
This session explains how solutions desired by such IT/Internet/Silicon Valley etc companies can look like, how they may differ from the more “classical” consumers of machine learning and analytics, and the arising challenges that current and future HPC development may have to cope with.
Learning Graphs Representations Using Recurrent Graph Convolution Networks Fo...Yam Peleg
Recently, number of papers re-visited this problem of generalizing neural networks to work on arbitrarily structured graphs, some achieving promising results in domains that have previously been dominated by other shallower algorithms. While Graph convolutions are generalization of spacial convolutions, and easiest to define in spectral domain, General Fourier transform used to represent them scales poorly with size of data. Therefore, first order approximation in Fourier-domain used to obtain efficient linear-time graph-CNNs. Those scales poorly with size of data. due to that, the expressiveness power of the proposed graph convolutional networks is severely impoverished. Another approach for learning graph representations requires the repeated application of contraction maps as propagation functions until node representations reach a stable fixed point. We combine those approaches and propose a recurrent version of Relational Graph Convolution networks, we then proceed to construct two models, Recurrent Variational Graph AutoEncoder and Recurrent Graph Convolution Regressor and show that for Ethereum Blockchain transaction graph we outperform the traditional Graph Convolution Network at predicting future movments of the corresponding tradable asset: Ether.
Machine learning for sensor Data AnalyticsMATLABISRAEL
במצגת זאת נראה כיצד עושים Machine Learning בסביבת MATLAB. נציג מספר יכולות ואפליקציות מובנות ההופכות את תהליך למידת המכונה ליעיל ומהיר יותר – כלים כמו ה-Classification Learner, ה-Regression Learner ו-Bayesian Optimization. בהסתמך על מידע המתקבל מחיישני סמארטפון, נבנה מערכת סיווג המזהה את הפעילות שמבצע המשתמש – הליכה, טיפוס במדרגות, שכיבה, וכו'
Our guide will provide you with a roadmap of the current situation, what this means for brands, and what you can do in the coming months to protect your brand’s vitality.
Creating great decks: The Origins, the "Why", and 12 Tips to Make Yours Better.Digital Surgeons
A big part of what we do is in the story we tell and how it’s presented. You’re probably thinking… decks, decks, and more decks. We hate em’, yet we love the good ones. There’s a certain formula that is used for every impactful story, speech, slide, and keynote. In this presentation we take a step back and really try to look at the elements of an impactful presentation. We've codified all of what goes into making a great deck, starting with the origins, the why, and ending with few tips to help elevate yours for whatever purposes they serve.
The Science of Story: How Brands Can Use Storytelling To Get More CustomersDigital Surgeons
Storytelling is not only an entertaining source for information, but a way to engage and humanize our messages that helps them stick. Our brains are wired for stories. Like a drug, we seek them out. Good stories create lasting emotional connections that persuade, educate, entertain, and convert consumers into brand loyalists.
Here’s another good reason to believe in the power of stories: You don't have a goddamn choice. We spend a third of our waking hours crafting stories, and the rest of the time consuming them. Our brains are always searching for stories. You need stories. You live your life around stories. Your life itself is a story. So, now find out how you can use them to better understand how brands and businesses can use storytelling to increase engagement and sales.
Great content is rooted in your audience's natural language, delivering a great content experience, search discoverability, and engaging storytelling. Quality, informative content that educates, persuades, entertains, or converts content consumers is the way forward for content creators hoping to engage with their audience.
L.E.S.S. Stands for:
Language
Experience
Search &
Storytelling
Unlock Your Organization Through Digital TransformationDigital Surgeons
Digital Transformation allows you to be disruptor, not the disrupted. See what you missed from our workshop at the Carnegie Mellon Engineering and Technology Innovation Management (ETIM) program’s 10th Anniversary Summit with senior leaders from academia and industry. Learn how to digitally optimize your business with principles of human-centered design that put the heart of the consumer at the center of business model innovation.
Digital Transformation
Design Thinking
Radical Candor: No BS, helping your team create better work.Digital Surgeons
Inspired by Google's Kim Scott, the Digital Surgeons team adapts Radical Candor to fit with their agile & innovative approach to designing the future of experiences.
Source: Candor, Inc.
http://www.radicalcandor.com/
Unlocking Creativity: How to Harness the Powers of Design, Art Direction & Cr...Digital Surgeons
Using gaming's concept of Progression, this presentation takes viewers on a journey that demystifies the roles and disciplines of Design, Art Direction, and Creative Direction – demonstrating how they can be mastered to take your creative work to the next level.
Fight for Yourself: How to Sell Your Ideas and Crush PresentationsDigital Surgeons
Don't let your blood, sweat, and pixels be overlooked, great creative doesn't sell itself.
Every presentation is a story, an opportunity to sell not just your work, but what people actually buy — YOU.
This presentation will walk viewers through three core aspects of winning at any presentation, Confidence, Comprehension, and Conviction.
These concepts, central to your work as a creative professional, are backed by science and bolstered by thoughts from some of the world’s leading creative professionals.
Better Twitch Broadcasting through Rapid Prototyping & Human Centered DesignDigital Surgeons
LIVESTREAMING IS BECOMING MAINSTREAM.
Human Centered Design is more than just another buzzword.
Players are now both the producers and the consumers of video content, creating new challenges and opportunities for publishers and brands.
The eSports industry is turning gaming into a lucrative spectator sport; over 200 million viewers in 2014 with over 3.7 billion hours watched.
The rise of Youtube Gaming, Periscope, and the $970m acquisition of Twitch show both the potential and popularity of streaming in the gaming community.
TWITCH HAS CHANGED THE GAME.
Twitch accounts for more than 43% of all live video-streaming traffic by volume.
BRANDS AND PUBLISHERS ARE STARTING TO SEE THE VALUE.
-Red Bull Twitch ’n Ride - the Red bull Twitch channel has 65,000+ followers
-Old Spice Nature Man - this Twitch campaign alone earned Old Spice over 32,000 followers
-Coca-Cola - partnering with League of Legends
Snickers - partnering with Twitch for their “You’re not You” campaign
WE FAIL FAST, EARLY, AND INEXPENSIVELY IN ORDER TO ARRIVE AT HUMAN CENTERED SOLUTIONS.
“GREAT DESIGN ALLOWS PEOPLE TO ACCOMPLISH THE SAME GOALS IN THE LEAST AMOUNT OF MOVES.”
DAN SAFFER
Author of Microinteractions: Designing with Details
eSports is changing the way we compete - http://esports.digitalsurgeons.com/
A brief primer for designers looking to improve their writing, learn about the historic intertwining of art directors and copywriters, and gain some tips on how to work collaboratively when marrying art and copy to create great work.
You’re not the expert. Your customers are, and who your customer is, is changing rapidly. Learn more about the digital consumer, how to bring new life to your customer experience, and inspire your team with workshop activities. Take a deeper look into the key drivers of your business, reinvigorate your customer experience, and gain insight from one of the newest inspiring entrepreneurs, who built his business around an out-of-the-ordinary customer experience. Why not create an experience that will leave your customers talking and sharing your brand with everyone? These musings were gathered after attending the Next Generation Customer Experience Conference in San Diego, March 2015.
Having a strong, unique and consistent Brand Voice is key to creating a successful brand across all marketing channels. This Brand Voice Toolkit will help you build a voice for your brand by first introducing the concept of Brand Voice and why it is imperative for a brand to be recognizable, identifiable, and relatable.
Your Brand Voice Toolkit should contain:
1. Brand Character + Personification
2. Brand Personality
3. Defined Vocabulary
4. Words Your Brand Says + Doesn’t Say
5. Writing Samples
Learn what each of these tools are and how they can be used to craft your Brand Voice in this deck and even explore an example toolkit.
Design Thinking: The one thing that will transform the way you thinkDigital Surgeons
What's the one thing that will transform the way you think? Design Thinking. The startups, trailblazers, and business mavericks of our world have embraced this process as a means of zeroing in on true human-centered design.
Design Thinking is a methodology for innovators that taps into the two biggest skills needed in today’s modern workplace: critical thinking & problem solving.
Of course, if you ask 100 practitioners to define it, you’ll wind up with 101 definitions.
Pete Sena of Digital Surgeons believes that Design Thinking is a process for solving complex problems through observation and iteration. At its core, he describes it as a vehicle for solving human wants and needs.
Minds are like parachutes; they only function when open. Thomas Dewar was a Scottish whiskey distiller.
Communicating ideas or insights is often the hardest part of the design process. And PowerPoint and Excel spreadsheets are limited in their ability to do this. But the communication tools used in Design Thinking—maps, models, sketches, and stories—help to capture and express the information required to form and socialize meaning in a very straightforward, human way.
The Five things that all definitions of Design Thinking have in common:
1. Isolating and reframing the problem focused on the user.
2. Empathy. A design practitioner from IDEO, the popular design and innovation firm strapped a video camera to his head and it was only then that he recognized why the ceiling is such an important factor when working with hospital patients. As a patient you lay in bed and stare at it all day. It’s these little details and true empathy that can only be realized by putting oneself in the user’s shoes.
3. Approach things with an open mind and be willing to collaborate. Creativity with purpose is a team sport.
4. Curiosity. We have to harness our inner 5-year-old here and really be inquisitive explorers. Instead of seeing what would be or what should be, consider what COULD be.
5 - Commitment. Brainstorming is easy. It’s easy to want to start a business or solve a problem. Seeing it into market and making it successful is not for the faint of heart. We’ve all read about big “wins” (multi-billion dollar acquisitions like Instagram and WhatsApp). What we don’t read about are people like Tony Fadell and Matt Rogers, who work for years before becoming industry sensations.
Pete describes what he refers to as the “Wheel of Innovation” as a process that continuously focuses on framing, making, validating, and improving on your concept. Be it as small as a core feature in your product down to the business model and business idea itself.
Design is about form and function, not art.
What are the business benefits for Design Innovation?
IDEO started an idea revolution when they coined this phrase DESIGN THINKING. Organizations ranging from early-stage startups up to Fortune 50 organizations have capitalized on this iterative appr
How YouTube is Drastically Changing the Beauty IndustryDigital Surgeons
Marketers of cosmetics can no longer simply rely on the photoshopped models of billboards, lifestyle magazines, and urban murals to secure market share.
Explore our comprehensive data analysis project presentation on predicting product ad campaign performance. Learn how data-driven insights can optimize your marketing strategies and enhance campaign effectiveness. Perfect for professionals and students looking to understand the power of data analysis in advertising. for more details visit: https://bostoninstituteofanalytics.org/data-science-and-artificial-intelligence/
Data Centers - Striving Within A Narrow Range - Research Report - MCG - May 2...pchutichetpong
M Capital Group (“MCG”) expects to see demand and the changing evolution of supply, facilitated through institutional investment rotation out of offices and into work from home (“WFH”), while the ever-expanding need for data storage as global internet usage expands, with experts predicting 5.3 billion users by 2023. These market factors will be underpinned by technological changes, such as progressing cloud services and edge sites, allowing the industry to see strong expected annual growth of 13% over the next 4 years.
Whilst competitive headwinds remain, represented through the recent second bankruptcy filing of Sungard, which blames “COVID-19 and other macroeconomic trends including delayed customer spending decisions, insourcing and reductions in IT spending, energy inflation and reduction in demand for certain services”, the industry has seen key adjustments, where MCG believes that engineering cost management and technological innovation will be paramount to success.
MCG reports that the more favorable market conditions expected over the next few years, helped by the winding down of pandemic restrictions and a hybrid working environment will be driving market momentum forward. The continuous injection of capital by alternative investment firms, as well as the growing infrastructural investment from cloud service providers and social media companies, whose revenues are expected to grow over 3.6x larger by value in 2026, will likely help propel center provision and innovation. These factors paint a promising picture for the industry players that offset rising input costs and adapt to new technologies.
According to M Capital Group: “Specifically, the long-term cost-saving opportunities available from the rise of remote managing will likely aid value growth for the industry. Through margin optimization and further availability of capital for reinvestment, strong players will maintain their competitive foothold, while weaker players exit the market to balance supply and demand.”
Adjusting primitives for graph : SHORT REPORT / NOTESSubhajit Sahu
Graph algorithms, like PageRank Compressed Sparse Row (CSR) is an adjacency-list based graph representation that is
Multiply with different modes (map)
1. Performance of sequential execution based vs OpenMP based vector multiply.
2. Comparing various launch configs for CUDA based vector multiply.
Sum with different storage types (reduce)
1. Performance of vector element sum using float vs bfloat16 as the storage type.
Sum with different modes (reduce)
1. Performance of sequential execution based vs OpenMP based vector element sum.
2. Performance of memcpy vs in-place based CUDA based vector element sum.
3. Comparing various launch configs for CUDA based vector element sum (memcpy).
4. Comparing various launch configs for CUDA based vector element sum (in-place).
Sum with in-place strategies of CUDA mode (reduce)
1. Comparing various launch configs for CUDA based vector element sum (in-place).
2. M A C H I N E
L E A R N I N G
Algorithms
Machine learning is a discipline focused
on getting a computer to analyze data
without explicit instructions, and come up
with conclusions about that data.
9. M A C H I N E
L E A R N I N G
Algorithms
An algorithm is a step by step
description on how to calculate
an output from an input
10.
11. M A C H I N E
L E A R N I N G
Algorithms
y = f(x)
12. M A C H I N E
L E A R N I N G
Algorithms
y = 12-x
This is the algorithm
13. M A C H I N E
L E A R N I N G
Algorithms
y = 12-x
x = 6
y = 12-6
y = 6
input:
output:
14. M A C H I N E
L E A R N I N G
Algorithms
y = 12-x
x = 6
y = 12-6
y = 6
input:
output:
15. M A C H I N E
L E A R N I N G
Algorithms
y = x/2
x = 12
y = 6
let's try
algorithm
16. M A C H I N E
L E A R N I N G
Algorithms
y = 12-x
This is the original algorithm
17. M A C H I N E
L E A R N I N G
Algorithms
y = 24-x
x = 24
y = 24-6
y = 18
input:
output:
18. M A C H I N E
L E A R N I N G
Algorithms
y = x/2
x = 24
y = 6
let's try
algorithm
19. M A C H I N E
L E A R N I N G
Algorithms
y = x/2
x = 24
let's try
algorithm
y = 6
20. M A C H I N E
L E A R N I N G
Algorithms
let's try y = f(x)
21. M A C H I N E
L E A R N I N G
Algorithms
y = 12-x
This is the algorithm
22. M A C H I N E
L E A R N I N G
Algorithms
y = (6*4-6-6)-x
23. M A C H I N E
L E A R N I N G
Algorithms
x
y
input:
output:
24. M A C H I N E
L E A R N I N G
Algorithms
Supervised
Reinforcement
Unsupervised
25. M A C H I N E
L E A R N I N G
Algorithms
Supervised machine learning is
the most common. The goal is
to figure out the algorithm
between an input and output.
26. M A C H I N E
L E A R N I N G
Algorithms
Supervised machine learning
approaches two types of problems.
27. M A C H I N E
L E A R N I N G
Algorithms
Classification
Regression
| |
y = f(x)
Facial detection
Object recognition
Speech to text
Sentiment analysis
Spam filtering
Hardware failure
Health failure
Financial market shifts
Customer churn prediction
28. M A C H I N E
L E A R N I N G
Algorithms
Supervised
Reinforcement
Unsupervised
29. M A C H I N E
L E A R N I N G
Algorithms
Supervised Unsupervised
Boundary
Clusters
30. M A C H I N E
L E A R N I N G
Algorithms
The system has no y,
just many bits of x
(known output)
(known inputs)
33. M A C H I N E
L E A R N I N G
Algorithms
Unsupervised machine learning
takes arbitrary (unlabelled)
data and tries to find
trends and groups.
34. M A C H I N E
L E A R N I N G
Algorithms
This is commonly called
"clustering," e.g. finding
similarities in bits of data.
Clusters
35. M A C H I N E
L E A R N I N G
Algorithms
Inversely, it can also be
used to find anomalies.
Clusters
36. M A C H I N E
L E A R N I N G
Algorithms
Unsupervised machine learning
is far less common, but
represents the "future" of many
AI applications, since most data
in the world is "unlabelled."
37. M A C H I N E
L E A R N I N G
Algorithms
Unsupervised machine learning is
also used for
"Dimensionality Reduction,"
e.g. reducing the number of
columns in your data that aren't
unique.
38. M A C H I N E
L E A R N I N G
Algorithms
Supervised
Reinforcement
Unsupervised
39. M A C H I N E
L E A R N I N G
Algorithms
Reinforcement machine learning
uses a "reward system" to teach
a machine to make continuously
"rewarding decisions."
40. M A C H I N E
L E A R N I N G
Algorithms
interpreter
reward
agent
environment
state
action
41. M A C H I N E
L E A R N I N G
Algorithms
This is used in many things from
video games to self-driving cars.
42. M A C H I N E
L E A R N I N G
Algorithms
It's also similar to "recommender
systems," where a system tries to
find associated products, content,
etc that a user might like.
43. M A C H I N E
L E A R N I N G
Algorithms
Classification
Regression
Clustering
Dimensionality Reduction
Reinforcement Learning
Logistic Regression
Support Vector Machines (SVM)
Random Forest (RF)
Naive Bayes
Genetic Algorithms
Principle Component Analysis (PCA)
Linear Discriminant Analysis (LDA)
Autoencoders
Linear Regression
Polynomial Regression
Neural Networks
Regression Trees and Random Forests
K-Means
Linear Discriminant Analysis
Recommender Systems
K-Nearest Neighbor
Matrix Factorization
(Stochastic Gradient Descent,
Alternating Least Squares)
Association Rules (Apriori, Elcat)
Deep Neural Networks
Q-Learning
State-Action-Reward-State-Action (SARSA)
Deep Q Network (DQN)
Deep Deterministic Policy Gradient (DDPG)
45. M A C H I N E
L E A R N I N G
PROCESSES
Let's train a system to figure
out whether an alcohol is
🍷wine or 🍺 beer.
46. M A C H I N E
L E A R N I N G
All machine learning starts with
some form of "data."
PROCESSES
47. M A C H I N E
L E A R N I N G
🍺 🍷
Attribute 1: Color (as a wavelength of light)
Attribute 2: Alcohol by Volume (as a percentage)
PROCESSES
48. M A C H I N E
L E A R N I N G
Next, we go to the grocery store
and get beer and wine, to
gather data.
PROCESSES
49. M A C H I N E
L E A R N I N G
Color (nm) Alcohol % Beer or Wine?
610 5 Beer
599 13 Wine
693 14 Wine
PROCESSES
50. M A C H I N E
L E A R N I N G
We then get the data into format
& location suitable for machine
learning. This is called
data preparation.
PROCESSES
51. M A C H I N E
L E A R N I N G
1. Collect Data
2. Randomize Order
3. Visualize Data to look for
pre-existing patterns
4. Split data into "training" and
"performance testing" sets.
PROCESSES
52. M A C H I N E
L E A R N I N G
Next we choose a model. I'll talk
about this more later, for now,
let's use a simple one.
PROCESSES
53. M A C H I N E
L E A R N I N G
Then we move onto training.
(the bulk of the process)
PROCESSES
54. M A C H I N E
L E A R N I N G
0
5
10
15
20
550 575 600 625 650
PROCESSES
55. M A C H I N E
L E A R N I N G
y = m(x) + b
output slope input y-intercept
PROCESSES
56. M A C H I N E
L E A R N I N G
0
5
10
15
20
550 575 600 625 650
PROCESSES
57. M A C H I N E
L E A R N I N G
y = m(x) + b
output slope input y-intercept
Weight: Multiplied Value
Bias: Added to the end result
slope
y-intercept
PROCESSES
58. M A C H I N E
L E A R N I N G
We then tweak weights and
biases in the algorithm to be
more accurate.
PROCESSES
59. M A C H I N E
L E A R N I N G
training data
model prediction
test & update
weights & biases
PROCESSES
60. M A C H I N E
L E A R N I N G
Finally, we evaluate the results
and modify as needed, tuning
parameters where necessary
(like number of training loops).
PROCESSES
61. M A C H I N E
L E A R N I N G
Final result: a functional
machine learning model.
model prediction
Color: 660nm
ABV: 12% 🍷
PROCESSES
63. M A C H I N E
L E A R N I N G
TECHNIQUES
Feature Learning
The ability of a system to automatically
detect classifications in raw data.
64. M A C H I N E
L E A R N I N G
Sparse Dictionary Learning
Learning a more generic
representation of input data that
gets rid of noise and outliers.
TECHNIQUES
66. M A C H I N E
L E A R N I N G
Anomaly Detection
Identification of rare items, events
or observations which raise
suspicions by differing significantly
from the majority of the data.
TECHNIQUES
67. M A C H I N E
L E A R N I N G
Decision Trees
Determining a likelihood particular
outcome based on a set of
observations.
TECHNIQUES
68. M A C H I N E
L E A R N I N G
Your chances of survival were good if you were
(i) a female or (ii) a male younger than 9.5
years with less than 2.5 siblings.
Titanic Survival Decision Tree TECHNIQUES
69. M A C H I N E
L E A R N I N G
Association Rules
Discovers interesting relations
between variables in large databases
TECHNIQUES
70. M A C H I N E
L E A R N I N G
For example, the
{onions, potatoes} => {burger}
rule found in the sales data of a
supermarket would indicate that if a
customer buys onions and potatoes together,
they are likely to also buy hamburger meat.
TECHNIQUES
72. M A C H I N E
L E A R N I N G
MODELS
Artificial Neural Networks
A framework for many
different machine learning
algorithms to work
together and process
complex data inputs.
73. M A C H I N E
L E A R N I N G
MODELS
Support Vector Machines
Finds a way to
accurately split
classes of data,
before it is
processed further.
74. M A C H I N E
L E A R N I N G
MODELS
Bayesian Networks
Known as "belief" or "causal"
networks. They predict outputs with
multiple inputs, taking into account
how inputs affect each other.
75. M A C H I N E
L E A R N I N G
MODELS
Bayesian Networks
76. M A C H I N E
L E A R N I N G
MODELS
Genetic Algorithms
Algorithms that mimic the process
of natural selection. Similar to
reinforcement learning, but rely
on more biologically inspired
things like genetic crossover,
mutation, and selection.