This video is all about building a handwritten digit image classifier in Python in under 40 lines of code (not including spaces and comments). We’ll use the popular library TensorFlow to do this.
IBM Vice President of Open Technology Todd Moore shares thoughts on the coming banner year for open source, touching on Containers/Kubernetes/OpenShift, A.I. tie-ins such as Tensorflow, ONNX, and Pytorch, as well as other cloud native technologies such as Kubeflow. Look for the Java programming language to make waves in 2020 as well.
Artificial Intelligence is used everywhere. Everyday you see it in search engines, shopping recommendations, and digital assistants. It’s also used in fraud prevention, the medical industry uses it in analysis of radiological scans, and meteorologists use it to better predict extreme weather. But what is AI? The short answer is: AI is the science of making machines smart. AI’s use algorithms to solve problems. They can consume and analyze enormous amounts of data to learn to complete a particular task. This is called machine learning. One of the most popular machine learning libraries is Google’s TensorFlow. TensorFlow has several wrappers in several languages making it accessible to just about anyone. Here’s how it works. Most machine learning frameworks will follow the same basic steps: collect data, build models, train the network, evaluate the results, predict outcome. There are other machine learning frameworks out there, but the thing I like best about TensorFlow is that it is open source and has great community support.
In machine learning the goal of training is to create an accurate model that answers questions correctly, most of the time.
In order to train a model, we need correct data.
The quality and quantity of the data acquired will determine the quality of results.
We need to prepare the data, which includes randomizing. Also some of the data is earmarked for training and other data is used for evaluation.
Choosing a machine learning model is an important step and is pertinent to what is being evaluated.
The training phase is the bulk of the machine learning workflow.
The model usually does a poor job in the beginning, but we can then help it make better decisions by adjusting it over iterative phases to reach better conclusions.
At some point one needs to determine when the adjustments are adequate for one’s needs, since the adjustments can go on forever.
Machine learning is used for answering questions and making predictions using models and data rather than human judgement.
Tensorflow is an open source machine learning library for research and production and is a great place to experiment and learn.