Shulou(Shulou.com)06/02 Report--

This article mainly introduces "what are the advantages of artificial intelligence". In daily operation, I believe that many people have doubts about the advantages of artificial intelligence. The editor consulted all kinds of materials and sorted out simple and easy-to-use operation methods. I hope it will be helpful for you to answer the doubts about "what are the advantages of artificial intelligence?" Next, please follow the editor to study!

AI x Design artificial design intelligence

As a low-level ability, the purpose of artificial intelligence AI is not to replace the work of traditional designers, but to improve the ability and efficiency of enhanced design through the computing power (computing power) and rules of computers, and to turn machines into designers' assistants by making machines learn to design.

The breakthrough of design intelligence actually benefits from the proposal and popularization of AI algorithm framework (deep learning model). In recent years, design intelligence is mainly used in design semantic extraction, style recognition, style transfer, design evaluation recommendation, design confrontation generation and so on.

Take Chestnut, for example, Xiaomi CC9 launched a feature called "magic for the sky" when it was released. Users take a picture with a sky background, which can be transformed into sunny, cloudy, night and other different styles of sky.

For "changing days", from the point of view of the presentation model of interaction design, it is only an one-click switch of the "design style" of the client, but if we look at the implementation model, we first need to implement the semantic cutting of the picture, so that the machine can learn to distinguish what is the sky. Then, through the input of a large amount of different style image data, the machine will learn what style is, which will be applied to different algorithm models, such as # convolution neural network CNN# combined with # attention mechanism Attention Networks# to extract key features to realize style classification. Finally, you can use # to generate the network GAN# generation model and discrimination model to train and output the optimal target style pictures to complete the user's "magic change" operation.

There are many classic examples of design intelligence. As interaction design, the above algorithm model framework and implementation details may not be our focus. In fact, we need to pay attention to the ability and boundary of AI algorithm, so as to think about how AI affects product interaction and user behavior. If you want to master the capabilities and boundaries of the algorithm, you must not miss the following basic concepts.

The Beauty of algorithm

1. "what is the algorithm? How do you use the algorithm? "

Generally speaking, an algorithm is actually a mathematical formula, which is a limited and definite set of solutions or problem-solving steps. For the algorithm, the input An and output B must be fixed, and the algorithm is only responsible for the middle output logic. For example, in order to calculate addition, subtraction, multiplication and division, the teacher summed up the formula of multiplication; for example, in order to enable users to receive goods faster, the distribution system developed a path planning optimization algorithm.

In the scenario application, the algorithm must be based on a certain business scenario pain point, in order to solve a certain kind of business problem abstraction, without the business scenario algorithm will lose its meaning. While solving business problems, we need to ensure the accuracy and integrity of the solution, which are the two major evaluation indicators that algorithm students usually use when delivering the algorithm model: accuracy and recall (recall). To put it simply, the accuracy measures the proportion of the algorithm model to predict an object or event is accurate, while the recall rate evaluates the proportion that the algorithm model can predict in the object or event heap to be predicted. Accuracy priority is more accurate, and recall priority is more comprehensive. No algorithm can be accurate and complete at the same time. It can only be based on the balance of business scenario and priority, but more attention is paid to accuracy in most e-commerce recommendation scenarios.

2. "is the algorithm subject to supervision or not?"

Just mentioned that the algorithm is a set of solutions under agreed input and output, then the training and learning methods of logic black box will be divided into supervised learning and unsupervised learning.

Supervised learning means that we agree with the machine in advance what is An and what is B, and the machine trains and learns according to our standards; supervised learning requires that the input of the algorithm model must be pre-processed manually, that is, data marking. The classic supervised learning methods, including regression and classification, the regression equation in basic mathematics is actually a kind of supervised learning.

Unsupervised learning, as the name implies, we do not need to define good standards and rules, the algorithm will automatically train and learn through feature extraction. Unsupervised learning is generally realized by clustering algorithms.

3. Summary

Algorithm is to agree on the input and output of the problem-solving steps, to let the machine learn to judge, you can teach it hand-in-hand (supervised), or let it become self-taught (unsupervised).

Machine learning VS deep learning

If algorithms are problem-solving solutions, machine learning and deep learning are more like problem-solving ideas and problem-solving structures at different stages.

1. What is machine learning?

Machine learning, in popular terms, is constantly trying and feedback to get the machine to accomplish a task well. If the machine wants to take part in the college entrance examination and the goal is to take the first place, according to the train of thought of machine learning, the machine will answer all the history questions, learn the connection between the questions and the answers, and then constantly answer the questions repeatedly to improve the accuracy, and finally take the exam to get the first place.

In the above chestnut, there are actually several key elements: task T, performance measurement P, and experience E. Task T is the ultimate goal of machine learning or algorithm, that is, to get the first place in the exam. Performance metrics P is the Benchmark for us to optimize data structures or algorithm logic. Defining appropriate performance metrics is very important for the efficiency and improvement of machine learning. The final experience E actually refers to historical data or effect data, that is, historical questions and wrong questions.

In the field of e-commerce, the classical people-goods matching recommendation algorithm can be described as, according to the above definition, in order to more accurately recommend personalized goods (T) to different users, through historical recommendation data and user behavior data (E), the problem-solving process of continuously improving the accuracy of recommendation results (P).

Summary: machine learning is based on historical data, outputs prediction data through model learning and training, and then continuously optimizes and improves the prediction model based on feedback data.

2. What's the difference between deep learning and machine learning?

Going back to our example, there will be not only objective questions but also subjective questions in the college entrance examination. if we come to grade the composition this time, it is very difficult for traditional machine learning to analyze according to the human way of thinking and writing logic. The emergence of deep learning enables machines to "think like human beings". Through the establishment of bionic neural networks, machines can learn and understand the connections between complex things, and optimize their own network structure by perceiving the outside world.

In other words, the deep learning model is actually a kind of machine learning, but the learning method has become, through the construction of a multi-layer perceptual structure of human-like brain neural network structure, as well as the combination and vectorization of low-dimensional features, to mine the characteristics of sample data, so as to establish the relationship between data and results.

To better understand deep learning and neural networks, we must first understand how the human brain perceives and understands information (data). The middle network diagram is an abstract representation of the structure of the human brain.

Each circle represents a neuron, and each arrow represents a set of synapses. Information enters from the left input layer, propagates through several hidden layers and synapses with different weights, and is output from the right output layer. Synapses with different weights filter or enhance information of different priorities to guide people's actions.

Based on the structure of human brain neural network, the deep learning model first converts high-dimensional sparse features into low-dimensional dense features by Embedding vectorization, then retains important features through the hidden layer, and then determines whether the training error meets the output requirements by the loss function Loss of the output layer, and finally completes the model output.

Compared with machine learning, the biggest breakthrough of deep learning lies in data feature mining, that is, the process of feature vectorization, so deep learning frameworks are generally used in complex unstructured data scenarios, such as speech recognition ASR, text processing NLP, image recognition CV and so on.

Summary: machine learning is a method to realize artificial intelligence, and deep learning is a technology for machine learning.

Product interaction and algorithm boundary

Algorithms and machine learning are actually solutions to problems. The effect of the algorithm depends to a large extent on the abstraction and transformation of the previous business problems.

Therefore, for products or interactive students, the key is to understand the technical capability boundary of the algorithm, based on the carding of business scenarios and requirements, to transform the business problem into an algorithm problem, and to find the most suitable and efficient algorithm framework to achieve business goals. At the same time, in the process of product interaction design, the educational cost of human-computer interaction should be reduced as much as possible to make the performance model close to the user psychological model.

At this point, the study of "what are the advantages of artificial intelligence" is over. I hope to be able to solve your doubts. The collocation of theory and practice can better help you learn, go and try it! If you want to continue to learn more related knowledge, please continue to follow the website, the editor will continue to work hard to bring you more practical articles!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.