The difference and use of machine learning and deep learning and the comparison of use cases

Today, the application of artificial intelligence is more and more extensive. The terms machine learning and deep learning also appear, and machine learning and deep learning are not exclusive or exclusive. Deep learning is a subset of machine learning, both of which are subsets of artificial intelligence (AI).

The difference and use of machine learning and deep learning and the comparison of use cases

Today, the application of artificial intelligence is more and more extensive. The terms machine learning and deep learning also appear, and machine learning and deep learning are not exclusive or exclusive. Deep learning is a subset of machine learning, both of which are subsets of artificial intelligence (AI). However, there are still many confusions in the definition and use cases of machine learning and deep learning, so I will clarify here:

• Artificial intelligence (AI) is the study of simulating and mimicking human intelligence behavior in computer systems and machines.

• Machine learning is a sub-area of ​​artificial intelligence that uses algorithms to apply the concept of artificial intelligence to computing systems. Identify by computer and take action based on data patterns. Over time, machine learning has no clear programming to improve its accuracy. Machine learning lags behind predictive coding, clustering, and visual heat map analysis.

• Deep learning is a subfield of machine learning and another name for artificial neural networks. Deep learning computer networks simulate how the human brain perceives, organizes, and makes decisions from data entry.

Machine learning and deep learning

In fact, the idea of ​​machine learning and deep learning ignores the fact that "deep learning is a subset of machine learning." People are more likely to use machine learning in their applications than deep learning. Deep learning is still a developing technology, and it can be expensive if deployed. However, some products are currently available, and deep learning will become more and more common over time.

Let's look at the differences and usage between the two:

Machine learning

As a subset of artificial intelligence, machine learning uses algorithms to parse data, learn from results, and use machine learning to make decisions or predictions. Examples include clustering, Bayesian networks, and visual data mapping. For example, in e-discovery and compliance surveys, heat maps and visual clusters present graphical search results to workers who can use the results to delve into other ambiguous data.

Machine learning techniques fall into two categories: supervised machine learning and unsupervised machine learning. Supervised learning depends on artificially generated seed sets and teaches software how to define data. A good example is predictive coding, which refers to a seed set that matches the data pattern to the relevant percentage. Over time, the predictive coding tool will learn from ongoing reviewer feedback.

Unsupervised machine learning relies on identifying the patterns contained in the data and comparing them to other data or search queries. Machine learning algorithms learn over time as data sets grow and more patterns emerge. Unsupervised machine learning includes clustering, concept search, and approximate deduplication.

For example, clustering matches similar text and metadata between documents and presents the data in a visual cluster. Conceptual search extends text-based queries by identifying and matching concepts. By approximating the deduplication of similar data and excluding the document based on similarity, in the email thread, the orphaned email is linked to the corresponding thread. Each of these analyses learns from its actions to improve performance and accuracy.

Machine learning infrastructure varies widely, with a single system enabling limited cluster or network traffic reporting, while large systems contain dozens of servers and massively parallel processing (MPP) architectures to handle massive amounts of data across multiple data sources.

Deep learning

Deep learning (also known as artificial neural networks) is based on all machine learning algorithms. However, it does not use task-specific algorithms such as data classification. Instead, it simulates human brain structure and function by identifying representative data from unstructured inputs and outputting accurate actions and decisions.

Deep learning can be supervised or unsupervised, which means that large neural networks can accept tagged input, but don't need it. The learning program teaches the neural network how to build different processing layers, but when the network processes the input, they create their own layers based on the data input and output. This level of deep learning allows the neural network to automatically extract features from the raw data without the need for additional manual input.

A neural network consists of a number of simply connected processors (called neurons) that are mathematical functions created to mimic neurons in the human brain. These artificial neurons make up the unit of the neural network. Simply put, each neuron receives two or more inputs, processes them, and outputs a result. Some neurons receive input from external sensors, while others are activated by inputs from other active neurons. Neurons may activate additional neurons or may affect the external environment by triggering actions. All activities are done in the own hidden layer, and each successive layer is input to the output of the previous layer.

In fact, neural networks take in unstructured data: sound, text, video, and images. The network divides the data into blocks of data and sends them to individual neurons and layers for processing. Once this discrete processing is complete, the network produces the final output layer.

Large-scale scalability is the key to neural networks. The performance of a neural network depends on how much data it can ingest, train, and process. The more data, the better the results. This is another place that is different from machine learning, and its algorithms usually remain stable at a certain level. Deep learning only limits its performance through its computing resources. Therefore, the "deep" part of the neural network: the more computing resources, the deeper the hierarchy, the wider the output. Although deep learning is not so fast and easy, the lower computational processing power has revolutionized research and development.

Common use cases for machine learning and deep learning

It is important to remember that the use cases for machine learning have already appeared on the market. The use cases for deep learning are mainly development goals at the current stage, and their commercialization is limited. Some use cases are similar: the difference is that neural networks can grow to near-infinite learning and output scale. Machine learning is more constrained and suited to specific real-world computing tasks. Also keep in mind that deep learning and machine learning are not mutually exclusive.

Comparison of machine learning and deep learning use cases

·Marketing:

Machine Learning: Emotional analysis of compliance emails and social media uses text prompts to alert emotional states.

Deep learning: Identify real-time emotions from photos and videos through sentiment analysis. Its applications include providing dynamic content or visual displays based on the emotional response of human observers.

·self-driving car

Machine learning: driverless cars based on sensor information

Deep learning: Driverless cars based on visual pattern recognition, for example, can immediately identify the difference between fire hydrants and pedestrians. Deep learning can also detect car defects by detecting the waveform of the engine's sound.

·Monitoring analysis

Machine Learning: The Monitoring Analysis System is rule-based and analyzes feeds through user-generated definitions and rules.

Deep learning: These systems are based on behavioral analysis. It observes detailed visual cues and teaches itself which behaviors are the normal behavior of the observed object. The networked deep learning system also recognizes identical faces from photos and videos of different time and place.

·life sciences

Machine Learning: A medical user program algorithm is used to identify variables in a diabetic population.

Deep learning: predicting the risk factors of individual patients. Cancer markers in MRI scans are identified by identifying visual markers.

·Speech Recognition

Machine Learning: Machine learning enables speech recognition to learn from users over time. This process is an intensive training that achieves an average of 95% accuracy.

Deep learning: The neural network processes billions of dictation audio clips, increasing the accuracy of speech recognition to nearly 100% while reducing training time. Speech recognition also classifies the original audio by keywords and themes and identifies the speaker, which has a wide impact on the development of audio surveillance technology.

·entertainment show

Machine Learning: Large CGI games can use machine learning to help automate intensive visual effects, such as exploding digital or visually huge space explosions.

Deep learning: Deep learning enables media and games to dynamically draw animations in response to user input, actions, and/or expressions.

Looking to the future of machine learning and deep learning

People do not find examples of deep learning/artificial neural networks in their daily lives. They generally require a large amount of tagged data for supervised learning, or a large amount of unstructured data for unsupervised learning. Deep learning technology developers spend a lot of time marking and inputting data to neural networks, or need to enter millions of unstructured objects to achieve unsupervised learning.

In today's data-intensive world, having enough data is not an issue. Marking enough data or introducing enough unlabeled data into a neural network is a challenge for businesses. Despite the increasing processing power of enterprises and the decline in hardware prices, intensive computing still requires a large amount of investment in systems and support.

Despite this, deep learning has good use cases in many different vertical areas of business. Industry giants like Google and Facebook are investing in deep learning research.

ECU Connector Section

ECU Connector Section.

ECU-MEM-CONNECTOR is used to connect the ECU-MEM to the DIAG-LAYER. This part of ODX also links the ECU-MEM-objects with DIAG-COMMS e.g. used to read idents, download or upload the data. It allows an ECU to be flashed using ODX.

There are nearly 100 kinds of connectors used in general automobiles, and hundreds of connectors are used in a single vehicle model. With the increasing demand for automobile safety, environmental protection, comfort and intelligence, the application of automotive electronic products is increasing, which will make the number of Automotive Connector applications increase.



Four basic structural components of automobile connector

1. Contact

It is the core part of automobile connector to complete the electrical connection function. Generally, the contact pair is composed of positive contact and negative contact, and the electrical connection is completed by inserting the female and male contacts. The positive contact is a rigid part, and its shape is cylindrical (round pin), square cylinder (square pin) or flat (insert piece). The positive contact is generally made of brass and phosphor bronze.




The female contact, namely jack, is the key part of the contact pair. It relies on the elastic structure to produce elastic force due to elastic deformation when it is inserted with the pin, and forms a close contact with the positive contact to complete the connection. There are many kinds of Jack structures, such as cylinder type (split slot, shrink mouth), tuning fork type, cantilever beam type (longitudinal slot, 9-shaped), box type (square socket) and hyperboloid linear spring jack.



2. Shell

Also known as the shell, it is the cover of automobile connector, which provides mechanical protection for the built-in insulation mounting plate and pin, and provides the alignment of plug and socket when inserting, so as to fix the connector to the equipment.



3. Insulator

Insulator is also called automobile connector base or insert. Its function is to arrange the contact according to the required position and spacing, and to ensure the insulation performance between contacts and between contact and shell. Good insulation resistance, voltage resistance and machinability are the basic requirements for selecting insulating materials to be processed into insulators.



4. Annex

Accessories are divided into structural accessories and installation accessories. Structural accessories such as snap ring, locating key, locating pin, guide pin, connecting ring, cable clamp, sealing ring, sealing gasket, etc. Install accessories such as screws, nuts, screws, coils, etc. Most accessories have standard parts and general parts. It is these four basic structural components that enable the automotive connector to act as a bridge and operate stably.



Application characteristics of automotive connector



From the use purpose of the automobile connector, in order to ensure the better driving of the car, we can divide the reliability of the connector into the sealing property in use and the fireproof performance in the driving of the car. In addition, in the driving of the car, the connector can also show the shielding performance and temperature control performance. In general, when discussing the sealing property of automobile connector, the sealing performance of automobile connector is not only for water.



In this field, the current international management standard is IP67, which is also the highest level in the automotive closure industry. Although the waterproof requirements are different in different parts of the car, many automobile manufacturers will choose IP67 to ensure the airtight performance of their car connectors.



In the driving of the car, power supply is a very important energy source, not only related to the normal work of the engine, but also often involved in the use of power supply in the process of driving. Therefore, in the automotive power system, the automotive connector has different division, between the male connector and the equipment, between the male connector and the cable, between the male and female connector, and between the female connector and cable, there are certain sealing measures that can be implemented.



Sealing ring is a common tool in the sealing performance of automotive connector. This tool can not only achieve the effect of fixing between different hole positions, but also achieve the sealing effect. It can not only ensure the stable operation of automobile equipment, but also ensure the waterproof performance of automobile equipment in work. Most of the sealing rings are made of silicone rubber, which is made of liquid silicon and solid silicon after certain chemical changes.



Now the car is in use, electronic circuit technology is an important aspect of the automotive industry, not only in the driver's entertainment, but also in the driver's driving car control system, electronic circuit technology plays an important role in the stable work of the car. In order to ensure the stable work of electronic circuit technology, people have applied a lot of shielding technology in automobile production.



These shielding technologies not only play a protective role in the electronic circuit of the automobile, but also have the ability of anti-interference and anti radiation in the parts of the car. In addition, they can also play a protective role for the stable work of the automobile connector. These can be divided into two types of external shielding and internal shielding.



When using external shielding to protect automobile connector, two identical shielding shells are usually assembled together to form a shielding layer, and the length of the shielding layer can cover the whole length of the connector. On the shielding layer shell, there should be enough locking plate structure to ensure the reliable installation of the shielding layer. In addition, the shielding materials used should not only be treated by electroplating, but also be treated to prevent chemical corrosion.



The future trend of automotive connector

At present, the global automotive connector industry accounts for about 15% of the connector industry. In the future, it is expected to take a larger proportion under the drive of automotive electronic products. In terms of product cost structure, the average cost of connectors used in each car in China is only a few hundred yuan, compared with that of foreign countries

ECU Connector Section

ShenZhen Antenk Electronics Co,Ltd , https://www.antenkwire.com