All in-game materials, links, and videos from the while True: learn() game, simulator of a machine learning specialist.
Courses
Machine Learning | Coursera
Machine learning is the science of getting computers to act without being explicitly programmed. In the past decade…www.coursera.org
MIT 6.S094: Deep Learning for Self-Driving Cars
This class is an introduction to the practice of deep learning through the applied theme of building a self-driving…selfdrivingcars.mit.edu
Spinning Up in Deep RL
We're releasing Spinning Up in Deep RL, an educational resource designed to let anyone learn to become a skilled…blog.openai.com
Machine Learning Crash Course | Google Developers
An intensive, practical 20-hour introduction to machine learning fundamentals, with companion TensorFlow exercises.developers.google.com
Open Machine Learning Course mlcourse.ai
mlcourse.ai is an open Machine Learning course by OpenDataScience. The course is designed to perfectly balance theory…mlcourse.ai
Introduction to Deep Learning | Coursera
The goal of this course is to give learners basic understanding of modern neural networks and their applications in…www.coursera.org
Technologies
ARMA
ARMA is a special machine learning model, specifically adapted to work with time-dependent numerical data. It can predict future indicators basing on a set of previous values. It is often used in economics, for example, to predict exchange rates.
ARMA Properties and Examples - Week 5: Akaike Information Criterion (AIC), Mixed Models, Integrated…
Video created by The State University of New York for the course "Practical Time Series Analysis". In Week 5, we start…ru.coursera.org
Isolation forest
Isolation forest is one of the simple algorithms for detecting anomalies. Same as the Decision tree and the Random forest, Isolation forest distributes the incoming elements into the leaves of the trees. The fewer ‘questions’ it took to separate an element from the rest, the more anomalous it is considered to be.
Anomaly Detection using the "Isolation Forest" algorithm
Anomaly detection can provide clues about an outlying minority class in your data: hackers in a set of network events…cds.cern.ch
SIFT
Introduction to SIFT (Scale-Invariant Feature Transform) - OpenCV-Python Tutorials 1 documentation
In last couple of chapters, we saw some corner detectors like Harris etc. They are rotation-invariant, which means…opencv-python-tutroals.readthedocs.io
Rosenblatt Perceptron
Perceptron - Wikipedia
In machine learning, the perceptron is an algorithm for supervised learning of binary classifiers. A binary classifier…en.wikipedia.org
Decision Tree
A Complete Tutorial on Tree Based Modeling from Scratch (in R & Python)
Introduction Tree based learning algorithms are considered to be one of the best and mostly used supervised learning…www.analyticsvidhya.com
Stochastic Gradient Descent
http://ruder.io/optimizing-gradient-descent/
Gradient Descent Intuition - Linear Regression with One Variable | Coursera
Video created by Stanford University for the course "Machine Learning". Linear regression predicts a real-valued output…ru.coursera.org
Gradient Descent
http://ruder.io/optimizing-gradient-descent/
Gradient Descent Intuition - Linear Regression with One Variable | Coursera
Video created by Stanford University for the course "Machine Learning". Linear regression predicts a real-valued output…ru.coursera.org
RNN
Sequence Models | Coursera
Sequence Models from deeplearning.ai. This course will teach you how to build models for natural language, audio, and…www.coursera.org
A Beginner's Guide to LSTMs and Recurrent Neural Networks
Data can only be understood backwards; but it must be lived forwards. - Søren Kierkegaard, Journals Contents Actually…deeplearning4j.org
Random Forest
The Random Forest Algorithm
Random Forest is a flexible, easy to use machine learning algorithm that produces, even without hyper-parameter tuning…towardsdatascience.com
Genetic Algorithms
Introduction to Genetic Algorithm & their application in data science
Introduction Few days back, I started working on a practice problem - Big Mart Sales. After applying some simple models…www.analyticsvidhya.com
Expert Systems
Artificial Intelligence Expert Systems
Artificial Intelligence Expert Systems - Learning Artificial Intelligence in simple and easy steps starting from basic…www.tutorialspoint.com
Deep News Issues
Deep neural nets
There are neural networks that solve the so-called “xor problem” (an unsolvable problem for Rosenblatt’s perceptron). They have one “hidden” layer between the input and output layers. Such networks are called “non-deep”.
If the network has more than one hidden layer, it is called “deep”. The deeper the network, the more complex dependencies it is able to recognize. For example, the current image recognition networks include several thousand hidden layers. Amazing, isn’t it?
World champion vs genetic algorithm
One of the greatest chess games of all time is, without a doubt, the battle between Garry Kasparov and the Deep Blue supercomputer by IBM i1997. The first game was very difficult and tense, Kasparov had an advantage at first, but starting from move 44, many believe that he ceased to understand the logic of the computer, and in the end, lost the entire game.
Neural networks architecture
In neural nets, programmers use a vast range of different neurons. Inside each neuron, a math function is located.
For example, Rosenblatt’s neuron used a linear function. Simple neural net, which can solve “xor problem”, uses “relu” function.
These functions are called “activators”. For different tasks people use different activators which will be more suitable for certain cases. To solve more difficult issues, a combination of neurons with different activators is used.
Non-neural artificial intelligence
Mankind has always been trying to create an artificial intelligence. Before the invention of neural networks, people used expert systems.
An expert system is a deterministic algorithm, which reproduces decisions of a real person.
An example this is “ELIZA”, the very first chat bot in the world. It was created in 1966, and “talked” with the patient using questions similar to real psychotherapist’s.
It worked horribly.
The Darkest Years of Machine Learning
In 1969 Seymour Papert & Marvin Lee Minsky wrote a book called “Perceptrons”. In this book they talk about math constraints of the first perceptrons. (XOR problem).
This book has shifted the scientific interest and funds distribution from the US Government organizations, slowing down the progression of machine learning for almost 30 years.
The expanded version of the book was released in 1987, containing the chapters that disproved the statements from the critical remarks made since 1969.
A Book by a Recurrent Neural Network
In summer of 2017, a programmer named Zack Tutt created an RNN (Recurrent Neural Network), which predicts the events of george R. R. Martin’s sixth book, “The Winds of Winter”.
The neural network wrote the end of the story by analyzing the plot and style of the previous books. Chapters written by that program confirmed many fan theories.
Rosenblatt’s Perceptron
First perceptron (machine neuron) was invented by Frank Rosenblatt in 1957
Frank created MARK-1 in 1960
MARK-1 is a perceptron machine that can recognize English letters by their shapes
Vanishing gradient
The problem of ‘Vanishing gradient’ is a big problem in the process of training deep neural networks, especially recurrent neural ones. Passing through a number of layers of the network, the gradient becomes so small that it stops changing the weights of the network and the network stops learning.
There are several ways to solve this problem. One of those is using shortcuts (Gates) to pull the gradient through the network layer.
As a result of applying this method, the LSTM layer (Long Short Term Memory) was made, which allows realizing short-term and long-term memory of the network.
Watch
Educational videos will quickly teach you the basics.
Siraj Rival is making a kind of YouTube show about machine learning. He never answered to our email about while True: learn() but please give him a try.
Learn Python
Learn to program in a language used by ML professionals around the world
Learn to Program: The Fundamentals | Coursera
Behind every mouse click and touch-screen tap, there is a computer program that makes things happen. This course…www.coursera.org
Applied Data Science | Coursera
Applied Data Science from IBM. This is an action-packed specialization is for data science enthusiasts who want to…www.coursera.org
Code and Practice
Develop your skills and take part in competitions in machine learning. Remember that for the victory in some of them pay money
Competitions | Kaggle
Edit descriptionwww.kaggle.com
Earn Money!
The average salary of a specialist in machine learning is 144,000 dollars according to indeed
But remember technology is only a tool to help you make good things.