Introduction to Deep-Learning
What Does Deep-Learning Mean?
Deep learning is a subset of machine learning which in-turn is a subset of artificial intelligence as shown in Fig.1, . Deep learning (DL) differs from machine learning (ML) — also called shallow learning — that it offered better performance on many problems by utilizing a huge number of neural networks, huge datasets and have the ability to accelerate computing on GPUs. Most of nowadays AI advancements, such as detecting spam in emails, forecasting stock prices, recognizing images in a picture, diagnosing illnesses and self-driving cars are due to the great progress and power of deep learning.
Is there another difference?
Yes, another major difference between ML and DL that in DL, solving a problem becomes more easy because it completely automates what used to be the most crucial step in a machine-learning workflow: feature engineering. As what happens in using DL to classify images, we will not extract features from pictures then introduce them to an algorithm like what we are doing in ML, instead DL will do that itself as shown in Fig.2.
So, When to use?
Machine learning algorithms specifically, random forest and gradient boosting (or XGBoost) are used for problems where structured data is available, whereas deep learning is used for perceptual problems such as image classification.
What is Neural Network?
Artificial neural network is the building block of DL and it represents a trial to mimic the human brain in performing computations, and taking decisions and this is being built on the concept of learning (deep learning). Neural networks are built from small units known as preceptron or neuron. These preceptrons are interconnected with what so-called set of synapses (nerve endings), or simply connecting-links (we will handle it and how it works in another tutorial).
GPUs Rule in Deep Learning
May this title answer the question of why deep learning becomes more famous these days?.
Because of that, typical deep-learning model used to solve a computer vision or speech recognition problem may require orders of magnitude more the computational power than what your traditional PC’s CPU can deliver. This approach stimulates some of big companies like NVIDIA and AMD to employ their researches in GPU developments not for “the more realistic games” only, but also to be used in modelling the deep neural networks.
NVIDIA launched CUDA compute platform, a programming interface for its line of GPUs.
AMD also develped “Radeon Instinct™ GPU-Powered Machine Learning Solutions”.
Google revealed its tensor processing unit ( TPU ) project: a new chip design developed from the ground up to run deep neural networks. Colaboratory, or “Colab” for short, is a product from Google Research. Colab allows anybody to write and execute arbitrary python code through the browser, and is especially well suited to machine learning, data analysis and education. More technically, Colab is a hosted Jupyter notebook service that requires no setup to use, while providing free access to computing resources including GPUs (for more info. here).
