I just started learning Python (a computer programming language) two years ago when I look an Artificial Intelligence Class. It was required to program to Tensor Flow, one of the top AI programming frameworks by Google.
Now I hear that Python is considered the number three in popularity among current computer programmers.
Cannon Beach at night
Sometimes a better strategy can win over more intelligence. For example, making a smarter wagon wheel is nice, but is it practical? Having a better strategy and a broader set of mental models can help discover something new or help to solve a mathematical or scientific puzzle.
To a person who uses hammers and nails to solve most problems, this person will use this model for almost any problem. Someone expert in mathematics will try to apply mathematics to solve a social issue, but this may not work.
These are some of my comments from an article: Mental Models: How to Train Your Brain to Think in New Ways. This article takes many examples from how Richard Feynman solved hard problems.
Cannon Beach, Oregon
Check out these links where there is some success in processing video with neural networks using a Convolutional Neural Network (CNN) followed by a Recursive Neural Network (RNN) using LSTM’s (Long Short-term Memory) modules.
Some machine learning or deep learning data sets can take a lot of CPU time, their doing a lot of matrix arithmetic operations. There is a lot of number crunching here.
If you want to write code for GPUs then I suggest you to try the following Udacity course. Introduction to Parallel Programming With CUDA
So the abbreviations are crazy, CPU stands for central processing unit, GPU means graphics processing unit. What is more crazy is that almost all computers have a CPU and a GPU. The CPU does almost all computing to run the various applications (programs) on the computer. The GPU runs the graphics card, but not all GPU’s are easy to use for math calculations beside its graphics.
Then Google is working on a variety of custom machine learning chips that are even faster.
just a snowblower, not a CPU or GPU.
Convolutional Neural Networks got inspired from the way our human visual cortex works (human eye) and classifies images like dog, cat, man, women, car, jet, or computer.
Convolutional Neural Networks are used for the classification of images.They do a great job because the training of the neural network breaks up the image into component parts and does classification based on various parts of the image. So a dog or cat or numerical number is learned by the various properties of the image. Here are some great links:
best tutorial: Convolutional Neural Networks (CNNs / ConvNets)
Google’s TensorFlow neural networks
Visualizing and Understanding Deep Neural Networks by Matt Zeiler
Visualizing and Understanding Convolutional Networks
Deep Learning Reading List
Google’s TensorFlow: Convolutional Neural Networks
wow, look at the ability for Machine Learning and Artificial Intelligence to learn your music rhythm and melody and play a duet on it.
Try it: Here! How does it work: Here!
It is here based on Google’s TensorFlow Machine Learning open source software. Check out this link about “Make Music and Art Using Machine Learning” at magenta.tensorflow.org Then look at Learning from A.I. Duet . Some real possibilities for expanding music and allowing software to help accompany the musician. Yes this will not make one a better musician, but it sure helps to make more music. The software is part of the work of the Google Brain team who work on Magenta project in collaboration with the Google Creative Lab team.
Now to take this to the next level, consider this note on
Style-conditioned music generation
Piano in the sunlight
With machine learning, Neural Networks are looking promising for deep learning. But there is the rise of Bayesians, which is researchers doing Artificial Intelligence (AI) through the scientific method starting with an hypothesis. There is lots of trial and error designing a Neural Network, lots of parameters to tweak as one tries to get the data to correlate. Can machine learning and neural networks help us design better neural networks?
Great video on Deep Learning and Multilayer Neural Networks, Deep Learning Summer School, Montreal 2015
Visual features: From Fourier to Gabor, Deep Learning Summer School, Montreal 2015
What is a MOOC? — a study course made available over the Internet for free to a large number of people
The Columbia River in the snow
To be creative, one must start somewhere. Creativity is bringing things together. But just brain-storm and consider things even outside the box, things that are impossible to achieve (so we think). Try a different mathematical solution, a different marketing or engineering solution. Success sometimes comes by just tweaking things just a little.
Consider the configuration of artificial intelligence and deep learning neural networks. There are so many ways to configure and initialize them. So many different types of neural nodes and ways to mathematically connect the nodes also called Perceptrons.
A person who never made a mistake never tried anything new — possibly said by Albert Einstein
Here is a great set of project results to consider for stimulating creativity in Machine Learning…
Higgs Boson Machine Learning Challenge
near the Columbia River in the snow
Maybe the top classes in Machine learning and Artificial Intelligence.
Intro to Artificial Intelligence Class Online – Udacity
Artificial Intelligence (AI) – Columbia University
Machine Learning and algorithms – Columbia University
For introductory material on machine learning and neural networks:
Great Google AI Platform: TensorFlow.org – Machine Intelligence
Also look at: Google Cloud Machine Intelligence and Google Cloud Machine Learning
Some AI platforms:
A Tour of Machine Learning Algorithms
Most platforms seem to use python programming language (or c++) for the interface.
- Data Analysis, Data Mining, Machine Learning and Mathematical Modeling are tools.
- Artificial Intelligence — automatic ways of reasoning.
- Machine Learning — turning data in to information and making decisions.
- Data Mining — extracting information from lots of data
- Artificial neural network (ANN)
- Pattern recognition