Machine Learning with TensorFlow

Deep learning is a subset of machine learning. In deep learning, machine learns underlying features in data using neural networks. This learning could be supervised or even unsupervised. Machine learning is a subset of artificial intelligence (AI) in which the machine has the ability to learn without having to be explicitly programmed. This is when we take sample data and train the machine on it so it makes predictions without having to be explicitly programmed. And finally, AI is supposed to be achieved when the machine can mimic human behavior.

Today I'd like to demonstrate how machine learning works. My example is based on TensorFlow.js, which is a WebGL accelerated, browser-based JavaScript library for developing machine learning models from scratch, running existing models in the browser and retraining existing models.

I was able to get this code from Tensorflowjs.org and when I run it you can see that we have one graph that has some data points. Chart no. 2 is an attempt to figure out the pattern without training TensorFlow on any data whatsoever and you can see that the pattern is completely off. If a human is asked to predict the pattern depicted on the chart, he should be able to create a line. That line is going to be very similar to the one we see in the third or final chart.

One of the misconceptions of beginners in machine learning is that there is no way to find out how the machine learning library arrives at the final solution, but that is actually not the case. You can see here, I made some changes in the code. Now, when I run this code you will see the transition of TensorFlow patterns from a completely random line to the final prediction, which is what we would expect human being would predict. If you like the video, do check out our article on Machine Learning and TensorFlow. If you have any training needs please visit TensorFlow training.