Intro to Deep Learning
By Kevin David Farinango
Deep learning is a type of machine learning that trains on a computer to perform tasks similar to human ones, such as recognizing speech, identifying images or making predictions. Instead of organizing the data to run through predefined equations, deep learning sets basic parameters about the data and trains the computer to learn on its own, it uses many successive layers of representations. Modern deep learning often involves tens or even hundreds of successive layers of representations, and all are automatically learned from exposure to training data. In deep learning, these layered representations are learned (almost always) through models called neural networks, structured in literal layers stacked on top of each other. The term neuronal red is a reference to neurobiology, although some of the core concepts in deep learning are developed in part by being inspired by our understanding of the brain, deep learning models are not models of the brain. For our purposes, deep learning is a mathematical framework for learning representations from data. The red transformation of the digit image into increasingly different representations of the original image and increasingly informative about the final result. You can think of a deep network as a multi-stage information distillation operation, where information passes through successive filters and increasingly purified sales (that is, useful with respect to some task).
Example convolutional neural network:
Convolutional Neural Networks (CNN): A convolutional neural network is a type of artificial neural network where neurons correspond to receptive fields in a manner very similar to neurons in the primary visual cortex (V1) of a biological brain. This type of network is a variation of a multilayer perceptron, however, because its application is carried out in two-dimensional matrices, they are very effective for artificial vision tasks, such as in the classification and segmentation of images, among other applications.
Convolution: In deep learning, a convolutional neural network (CNN, or ConvNet) is a class of deep neural networks, most commonly applied to analyze visual images.Pooling: Analysis and optimization of convolutional neural network architectures for grouping types in general.Convolution layer: Product and sum operations are performed between the starting layer and the n filters (or kernel) generated by a feature map. The extracted characteristics correspond to each possible location of the filter in the original image.The advantage is that the same filter (= neuron) serves to extract the same characteristic in any part of the input, thereby reducing the number of connections and the number of parameters to train compared to a multilayer network of total connection.
Pooling: The number of parameters is decreased by keeping the most common characteristics. The way to reduce parameters is done by extracting statistics such as the average or maximum of a fixed region of the characteristics map, by reducing characteristics the method loses precision but improves its compatibility.
Practical examples of deep learning
Virtual assistants: Alexa or Siri or Cortana, virtual assistants of online service providers use deep learning to help understand their speech and the language that humans use when they interact with them.
Google translate: Similarly, deep learning algorithms can automatically translate between languages. This can be powerful for travelers, businessmen and people in government.
Vision for driverless delivery trucks, drones and autonomous cars.The way in which an autonomous vehicle understands the realities of the road and how to respond to them, be it a stop sign, a ball in the street or another vehicle, is through deep learning algorithms. The more data the algorithms receive, the better they can act as humans in their information processing: knowing that a snow-capped signal remains a stop signal.
Facial recognition: Deep learning is being used for facial recognition not only for security purposes, but also to tag people in Facebook posts and we may be able to pay for items in a store simply by using our faces in the near future. The challenges for deep learning algorithms for facial recognition is knowing that it is the same person, even when the hairstyles have been changed, the beard has grown or shaved or if the image taken is poor due to poor lighting or obstruction .
Reference:
Versión en español: https://drive.google.com/file/d/1Ev3_w_GgPCWe7PRmgA4TUB6n6adZApOj/view?usp=sharing
https://livebook.manning.com/book/deep-learning-with-python/chapter-1/43
