Machine Learning
** House Liskov is not held liable for any misinformation provided. Shit complex.
What is Machine Learning?
Why do I care about it?
Weights of an algorithm
Supervised/Unsupervised Training
Neural Networks
Convolution
Real world Examples
What were covering:
* Machine Learning *
What People Think Machine Learning/Artificial Intelligence
What is Machine Learning?
Core-sub of Artificial Intelligence (AI)
A process that allows computers to learn & improvise on datasets/problems they haven't encountered before
Allows computer to be in a mode of self-learning.
WITHOUT being explicitly programmed.
finding patterns
solve new problems
Personal Interaction With Machine Learning
ALL THE PATTERNS!!!
What it looks like
So why is it so important?
To reiterate a key point, machine learning is used to identify working patterns in large sets of data. The more data, and the higher the quality, the better the results!
Machine learning is good at finding patterns that humans could likely identify as well. Data that has correlations. The difference is that a computer can do it at a more precise level provided it has the proper amount of reference data to learn from. Did I say a LOT of data?
Technology companies have been using machine learning to understand how users interact with their product. You may not realize how relevant and common this process is to they way you already use technology every day.
🤖🤖🤖🤖🤖
< Fucking feed me.
So what can machine learning be used to create? How does this even affect the average person?
Fo one, it helps creates better predictive
software features for the stuff you use everyday!
- Auto-completion for typing mid-sentence on smaller device such as smartphones
- Search phrase completion, like when using a search bar app like Google
- Optimization of search results that better match what the client is likely looking for
How about voice recognizing web entities like Amazon's Alexa, Apple's Siri, or other contenders in this market?
This more conversational type of user interface is intriguing to use, while also being more convenient for the user in many cases.
Descartes Labs is a more macro application that applies machine learning to giant satellite imagery data sets to analyze and predict crop yields. It helps forecast crop yields across the U.S. and the world, helping prepare areas for upcoming food shortages.
Machine learning is interesting because the derived patterns can be used in so many ways and on a huge variety of scales.
The enhanced patterns and associations that are generated can be used to help improve customer interaction with a product or website, as the customer is increasingly being given more natural and satisfying feedback as the product improves. All this is based on what similar users have "needed" in the past.
But since machine learning can be used to extrapolate from any type of data that can be digitalized, such as sound, visual input, vibration, heat, etc...and all on as large of a scale as allowed by the data source, amazing discoveries and technologies can be created and leveraged.
So...how does it work?
Basics of Machine-Learning
Basics of Machine-Learning
Basics of Machine-Learning
Basics of Machine-Learning
Basics of Machine-Learning
STEP 1:
Basics of Machine-Learning
STEP 2:
Basics of Machine-Learning
STEP 3:
Basics of Machine-Learning
STEP 3:
Basics of Machine-Learning
Basics of Machine-Learning
Neural Networks
By chaining lots of simple neurons together, we can model functions that are too complicated to be modeled by one single neuron. THUS neural network
We made a simple estimation function that takes in a set of inputs and multiplies them by weights to get an output. Call this simple function a neuron.
Neural networks can:
- Take in a ton of data over time to hone themselves (great for analyzing pictures)
- Adjust themselves! (good for changing weights)
Couple types O' Neural Networks
I never forget
Stateless (wtf happened?)
Problem:
if we move the image around in our image(on a diff background ect.) it will not be recognized by the neural network.
Convolution (network)
What is it?
- A way to train AI/ recognize images
- The object is the same regardless of where it appears in the picture
DEEP
Neural Network
What is Watson?
IBM's Watson supercomputer combines artificial intelligence (AI) and sophisticated analytical software for optimal performance as a “question answering” machine.
QA technology differs from a standard document search in that it takes a question or query posed in natural language and seeks to understand it in detail, returning a much more precise answer to the question.
Watson is currently being used in hundreds of applications, including recipe creation, early childhood education improvement, personalized cancer treatment information, and many more.
H&R Block
Watson will use natural language processing to parse data from the estimated 11 million tax returns that will be processed in 2017.
This data will then be related back to the 74,000-page federal tax code, helping to identify opportunities for additional deductions.
IBM Watson's cognitive computing platform can detect subtleties in language that can alert Twitter when someone is engaging in abusive behavior, and has the ability to parse data on huge repositories of data in a fraction of a second, much faster than any human, or team of humans for that matter, could.
What is Machine Learning?
Why do I care about it?
Weights of an algorithm
Supervised/Unsupervised Training
Neural Networks
Convolution
Real world Examples
WTF did I just hear?
deck
By iharvey
