In the past only science fiction dream of artificial intelligence (AI) is now a mainstream technology in our daily lives, and applied to the image and speech recognition, language translation, bot and predictive data analysis.
In this article, we will introduce AI and machine learning related terms and deep learning. At last, you should understand these terms, work under normal circumstances as well as more familiar terms such as Inception and YOLO (No, we're not talking about Leonardo DiCaprio movie or some internet memes).
Artificial Intelligence (AI) is a computer simulation of human intelligence.
Machine learning is a branch of AI, in which algorithms for learning in order to make future decisions or predict from the data.
Deep learning is a specific subset of the use of artificial neural networks (ANN) machine learning, artificial neural networks are inspired by the layered structure of the human brain.
The following figure shows more clearly AI, machine learning and deep learning interrelate.
Recently, machine learning has made significant progress and, more specifically, has made significant progress in depth learning. These improvements are due to the timely convergence of several factors, including the availability of large amounts of data, and the emergence of an efficient algorithm to increase computing power.
So how machine learning and deep learning work? In order to explain its most simple way, you have basically defined a model input (may be an image, audio, text, or numbers). The model will give a specific output according to the received input data (e.g., image classification, the detected object, etc.).
Its role is to provide you with a large number of sample data for the model to "train" it. With enough data samples, the model will be able to accurately "predict" the future output of any input. There are many well-known model for different tasks, for example, Inception image recognition model is a widely used, YOLO for object detection, FaceNet for facial recognition, TextCNN for text sentiment analysis, and so on.
Sounds simple, does not it? But make no mistake, the machine learning may not be rocket science, but it is no easy task. As an example, Inception model previously mentioned internal architecture of a network shown below, the idea over the years as more researchers climax:
It looks much more complicated than the big black box shown earlier! To determine the appropriate architecture, it requires a lot of experiments structured, and also require sufficient data for training and validation. Not just any data, as well as related data that properly cleaning and building
有几种工具和平台试图通过自动或可视化工具（例如Google Cloud AutoML，Microsoft Azure ML Studio，IBM Watson Studio，Lobe.ai等）简化AI模型训练，但事实是找到并训练适用于特定问题的深度学习模型需要数据科学家和机器学习专业人员，或者至少是有抱负的人员。为了更详细地了解机器学习，有许多在线课程可供选择，其中包括Andrew Ng在斯坦福大学备受推崇的机器学习课程。
TensorFlow是用于机器学习的领先开源AI库。TensorFlow由Google构建，旨在实施，培训，测试和部署深度学习神经网络。TensorFlow最初由研究Google Brain团队的研究人员和工程师开发供内部使用，于2015年作为开源项目发布.TensorFlow为多种语言提供API，包括Python，C，C ++，Go，Java，JavaScript和Swift。TensorFlow站点提供了许多入门教程。
