Home  ›  Deep Learning  ›  Machine Learning

Entry-level strategy: machine learning vs. deep learning

Your Ads Here

Entry-level strategy: machine learning vs. deep learning, deep learning tutorial, deep learning vs machine learning, deep learning ai, deep learning andrew ng, deep learning mit, deep learning pc, deep learning python, deep learning stanford, deep learning algorithms, deep learning applications, deep learning andrew ng course 1, deep learning attention, deep learning a crash course, deep learning andrew ng course 2, a deep learning framework for character motion synthesis and editing, a deep learning-based radar and camera sensor fusion architecture for object detection, a deep learning approach to grasping the invisible, a deep learning approach for generalized speech animation, a deep learning framework for financial time series using stacked autoencoders and lstm, a deep learning model for pediatric patient risk stratification, a deep learning framework for neuroscience, a deep learning approach for population estimation from satellite imagery, deep learning basics, deep learning build, deep learning bangla, deep learning book, deep learning basics with python tensorflow and keras, deep learning bangla tutorial, deep learning backpropagation, deep learning build 2020, deep learning course, deep learning computer, deep learning crash course, deep learning cmu, deep learning coursera, deep learning chatbot, deep learning chapter 2, deep learning computer vision, c language deep learning, c programming deep learning, deep learning c'est quoi, deep learning desktop, deep learning documentary, deep learning deepmind, deep learning deeplizard, deep learning demo, deep learning demystified, deep learning desktop 2020, deep learning drug discovery, deep learning explained, deep learning example, deep learning engineer, deep learning edureka, deep learning english, deep learning economics, deep learning experiments, deep learning gpu, machine learning e deep learning, differenza machine learning e deep learning, diferença machine learning e deep learning, deep learning for computer vision, deep learning fundamentals, deep learning full course, deep learning frameworks, deep learning from scratch, deep learning freecodecamp, deep learning for beginners, deep learning for coders, deep learning game, deep learning google, deep learning gpu benchmarks, deep learning genomics, deep learning gan, deep learning goodfellow, deep learning graphics card, deep learning hindi, deep learning hardware, deep learning hyperparameter tuning, deep learning harvard, deep learning history, deep learning hinton, deep learning hype, deep learning hello world, deep learning interview questions, deep learning in hindi, deep learning in r, deep learning introduction, deep learning in python, deep learning image processing, deep learning image classification, deep learning ian goodfellow, deep learning jobs, deep learning java, deep learning javascript, deep learning jeremy howard, deep learning julia, deep learning jupyter notebook, deep learning jetson nano, deep learning jarvis, deep learning for j, deep learning krish naik, deep learning keras, deep learning kelly howell, deep learning keras tutorial, deep learning keras tensorflow, deep learning kaggle, deep learning kumar, deep learning khan academy, deep learning p k biswas, deep learning lecture, deep learning laptop, deep learning lex fridman, deep learning laptop 2020, deep learning lstm, deep learning loss function, deep learning linear algebra, deep learning lex, deep learning music, deep learning machine, deep learning matlab, deep learning math, deep learning machine build, deep learning medical imaging, deep learning music generation, iitm deep learning, deep learning nptel, deep learning neural networks, deep learning nlp, deep learning nyu, deep learning neural networks and the future of ai, deep learning nvidia, deep learning ng, deep learning news, deep learning optimization, deep learning object detection, deep learning overview, deep learning on macbook pro, deep learning on aws, deep learning oxford, deep learning ocr, deep learning on windows, o que é deep learning, deep learning projects, deep learning pc build, deep learning pytorch, deep learning pc build 2020, deep learning python tutorial, deep learning projects in python, deep learning quantization, deep learning question answering, deep learning quick tutorial, deep learning que es, deep learning qgis, deep learning quiz coursera, deep learning quantum computing, deep learning quant trading, deep q learning tutorial, deep q learning pytorch, deep q learning tensorflow, deep q learning keras, deep q learning sentdex, double deep q learning, deep q learning tensorflow 2, dueling deep q learning, deep learning rig, deep learning research, deep learning recommender systems, deep learning robot, deep learning reinforcement learning, deep learning regularization, deep learning regression, deep learning roadmap, r cnn deep learning, deep learning in r studio, deep learning tutorial in r, deep learning keras r, deep learning specialization, deep learning server, deep learning super sampling, deep learning state of the art, deep learning sentdex, deep learning stock trading, deep learning simplified, deep learning tensorflow, deep learning transformers, deep learning tutorial andrew ng, deep learning trading, deep learning tutorial for beginners, deep learning tutorial python, deep learning time series forecasting, deep learning ucl, deep learning udacity, deep learning using python, deep learning use cases, deep learning using matlab, deep learning upscaling, deep learning uses, deep learning using keras, u net deep learning, deep learning video, deep learning voice, deep learning visualization, deep learning vs neural networks, deep learning voice cloning, deep learning vs machine learning vs neural network, deep learning video generation, gta v deep learning, titan v deep learning, deep learning with pytorch, deep learning with python, deep learning workstation, deep learning with tensorflow, deep learning with pytorch - free six week course, deep learning with python tensorflow and keras tutorial, deep learning workstation build, deep learning with pytorch a 60 minute blitz, deep learning x ucl, deep learning x ray, deep learning xor example, deep learning xor, 5700 xt deep learning, xception deep learning, xgboost vs deep learning, covid xray deep learning, x ray deep learning, jetson nano x deep learning 딥러닝, deep learning yann lecun, deep learning yolo, deep learning youtube, deep learning yoshua bengio, deep learning yolo object detection, deep learning youtube recommendations, deep learning yann lecun yoshua bengio & geoffrey hinton, deep learning youtube andrew ng, machine learning y deep learning, diferencia machine learning y deep learning, deep learning zero to hero, deep learning zero to gans, deep learning zero to all, deep learning a-z udemy, deep learning a-z, eth zurich deep learning, deep learning 101, deep learning 1x1 convolution, deep learning 1d convolution, kaggle deep learning 1, intro to deep learning #1, intro to deep learning #11, deep learning chapter 1, deep learning.ai course 1, lesson 1 deep learning 2019, course 1 deep learning, deep learning part 1, deep learning tutorial 1, deep learning 1, deep learning 2020, deep learning 2019, deep learning 2020 mit, deep learning 2080 ti, deep learning 20, deep learning 2048, deep learning 2d to 3d, deep learning 2.0 bengio, dota 2 deep learning, system 2 deep learning, deep learning part 2, deep learning course 2, deep learning week 2 assignment, deep learning starcraft 2, 2 minute papers deep learning, deep learning 3blue1brown, deep learning 3d, deep learning 3d modeling, deep learning 3d reconstruction, deep learning 37, deep learning 3 blue, deep learning 3d model generation, deep learning 34, 3 deep learning algorithms in under 5 minutes, lesson 3 deep learning 2019, deep learning chapter 3, deep learning course 3, kaggle deep learning 3, deep learning raspberry pi 3, deep learning 3, deep learning series ep 3, deep learning 4j, deep learning 4j tutorial, deep learning 4 java, deep mob learning skyfactory 4, deep learning course 4, raspberry pi 4 deep learning, deep learning chapter 4, 4 gpu deep learning build, unreal engine 4 deep learning, deep learning week 4 assignment, deep learning tutorial 4, deep learning на пальцах 4, deep learning 5 minutes engineering, deep learning 53, deep learning 5700 xt, deep learning in 5 minutes, deep learning course 5, deep learning chapter 5, deep learning ai course 5, gta 5 deep learning, deep learning book chapter 5, deep learning 6.s191, deep learning 6 c's, mit deep learning 6.s191, deep learning на пальцах 6, rtx 8000 deep learning
This article introduces the definition and application of machine learning and deep learning in plain and easy-to-understand language, as well as the differences in source data requirements, hardware support, feature engineering, problem-solving methods, execution time, and interpretability. Great enlightenment.
Wedge
Machine learning and deep learning are very popular now, and you will find that suddenly many people are talking about them.
This article will explain in detail the difference between deep learning and machine learning in simple and easy to understand language and examples, and introduce related uses.
Introduction to machine learning and deep learning

Machine learning

Tom Mitchell's definition of machine learning is widely quoted as follows:
For a certain type of task T and performance measurement P, if a computer program's performance measured by P on T continues to improve itself with experience E, then we call this computer program learning from experience E.
“A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E ”
The above abstract definition may confuse you, I believe the following simple examples will make you suddenly realize.
[Example 1 Predict weight based on height]
Suppose you want to create a system that predicts weight based on a person's height. The first step is to collect data. After collecting, draw the data distribution diagram as shown below. Each point in the figure represents a piece of data, the abscissa represents height, and the ordinate represents weight.
We can draw a simple straight line to predict weight based on height, such as:
Weight (in kg) = Height (in cm) – 100
If this straight line predicts height accurately, how to measure its performance? For example, the performance of the prediction model is measured by the difference between the predicted value and the true value. Of course, the more source data, the better the model effect. If the effect is not good, you can use other methods to improve the performance of the model, such as adding variables (such as gender) or changing the prediction line.
[Example 2 Storm Forecast System]
Suppose you want to build a storm prediction system. You have past storm data and weather data three months before these storms. So how to build a storm prediction system?
The first thing to do is to clean the data and find hidden patterns in the data, such as the conditions that caused the storm. We can model some conditions, such as whether the temperature is greater than 40 degrees Celsius, and whether the humidity is between 80 and 100, and then input these features into the model.
All you have to do is make full use of historical data and predict whether a storm will occur. In this example, the evaluation index is the correct prediction of the number of storms. We can repeat the prediction process many times and then return the performance results to the system.
Returning to the original definition of machine learning, we define the storm prediction system as follows: task T is to find the atmospheric conditions that caused the storm, performance P is the number of correct predictions after the model parameters are learned, and experience E is the iterative process of the system.

Deep learning

Deep learning actually appeared a long time ago, and with the hype in recent years, it has gradually become popular.
Deep learning is a special kind of machine learning that represents the real world as a nested hierarchical conceptual system (complex concepts defined by the connections between simpler concepts, from general abstraction to high-level abstract representation), so as to obtain powerful performance and flexibility.
Deep learning is a particular kind of machine learning that achieves great power and flexibility by learning to represent the world as nested hierarchy of concepts, with each concept defined in relation to simpler concepts, and more abstract representations computed in terms of less abstract ones.
[Example 1 Graphic detection]
Suppose we want to distinguish the rectangle from other graphics. The human eye first detects whether the graph has 4 edges (simple concept). If there are 4 edges, check whether they are connected, closed and vertical, and whether they are equal (the concept of nesting level). In fact, we decompose a complex task (rectangular recognition) into some simple low-level abstract tasks. Deep learning essentially does this on a larger scale.
[Example 2 Cat or Dog]
This case is to build a system that can identify whether the animal in the picture is a cat or a dog.
If you use machine learning to solve this problem, you must first define some characteristics, such as whether the animal has a beard and ears; if there are ears, whether the ears are pointed. Simply put, we have to define facial features, and then let the system recognize which are important features in animal classification. While deep learning will complete these tasks at once, deep learning will automatically find features important for classification tasks, and machine learning has to be manually specified.
The in-depth study workflow is as follows:
1. First find the border that is most relevant to the cat or dog in the picture;
2. Then find the combination of shape and boundary, such as whether you can find the beard and ears;
3. After continuous hierarchical recognition of complex concepts, it is possible to determine which features play an important role in identifying cats and dogs.
Comparison of machine learning and deep learning

Data dependence

The most important difference between deep learning and traditional machine learning is that its performance increases as the amount of data increases. If the data is small, the performance of the deep learning algorithm is not good. This is because the deep learning algorithm requires a large amount of data to understand the patterns contained therein. In this case, traditional machine learning that uses manually specified rules has the upper hand. 

Hardware support

Deep learning algorithms rely heavily on high-end machines, while traditional machine learning can run on low-end machines. Because deep learning requires a large number of matrix multiplication operations, and GPUs can effectively optimize these operations, GPUs have become an essential part of it.

Feature engineering

Feature engineering inputs domain knowledge into the feature extractor, reduces the complexity of the data, makes the patterns in the data more obvious to the learning algorithm, and obtains better results. In terms of time and professionalism, this process is expensive. In machine learning, most of the features used are designated by experts or determined by prior knowledge for each data field and data type. For example, features can be pixel value, shape, texture, location, and direction. The performance of most machine learning methods depends on the accuracy of identifying and extracting these features.
Deep learning algorithms try to learn high-level features from data. This is a distinctive part of deep learning and an important step beyond traditional machine learning. In deep learning, each problem is attributed to the development of new feature extractors. For example, convolutional neural networks learn low-level features such as edges and straight lines at the bottom, then facial features, and finally high-level features of human faces.

Problem solution

When using traditional machine learning methods to solve problems, they often adopt the strategy of breaking up into parts, solving them separately, and then combining the results to solve them. And deep learning advocates end-to-end model, input training data, directly output the final result, let the network learn how to extract key features.
For example, if you want to perform target detection, you need to identify the category of the target and indicate its position in the picture.
ypical machine learning methods divide this problem into two steps: target detection and target recognition. First, use border detection technology, such as grabcut, to scan the entire image to find all possible objects, and use target recognition algorithms such as HOG/SVM for these objects to identify related objects.
Deep learning methods deal with this problem in an end-to-end manner. For example, YOLO net can achieve target positioning and recognition through convolutional neural networks. That is, the original image is input to the convolutional neural network, and the position and category of the target in the image are directly output.

execution time

Generally, deep learning requires a long time to train, because many parameters in deep learning require much more than normal time training. ResNet takes about two weeks to complete training from scratch, while machine learning only requires training time ranging from a few seconds to a few hours. The time required for testing is exactly the opposite. Deep learning algorithms require very little time to run. However, compared with KNN (K Nearest Neighbor, a machine learning algorithm), the test time will increase as the amount of test data increases. However, not all machine learning algorithms take a long time, and some require very little testing time.

Interpretability

Assuming that you use deep learning to automatically score articles, you will find that the performance will be very good and close to the level of human scoring. But it cannot explain why such a score is given. During the operation, you can discover which nodes of the deep neural network are activated, but you don't know what these neurons are modeling and what each layer is doing, so you can't explain the result.
On the other hand, machine learning algorithms such as decision trees clearly explain the reasons for making choices at each step in accordance with the rules. Therefore, algorithms such as decision trees and linear/logistic regression are widely used in the industry due to their good interpretability.
Machine learning and deep learning application scenarios
The Wiki introduces some application areas of machine learning:
1. Computer vision: such as license plate number recognition, face recognition;
2. Information retrieval: such as search engine, text retrieval, image retrieval;
3. Marketing: automatic email marketing, target identification;
4. Medical diagnosis: cancer diagnosis, abnormal detection;
5. Natural language processing: semantic analysis, photo tagging;
6. Online advertising, etc.
The following figure summarizes the application areas of machine learning. In general, the application range is very wide.
Google is a well-known company in the industry that uses machine learning/deep learning. As shown in the figure below, Google applies deep learning to different products.
Instant test
In order to assess whether you really understand the difference between machine learning and deep learning, there will be a quick test here and you can submit your answers here. All you have to do is to use machine learning and deep learning to solve the following problems and decide which method is better.
[Scenario 1] Suppose you want to develop a self-driving car system. The system takes the raw data taken by the camera as input, and then predicts the direction and angle of the steering wheel.
[Scenario 2] Given a person's credit certificate and background information, evaluate whether a loan can be granted to him.
[Scene 3] Create a system to translate Russian text into Indian language.
Future trend
The previous summary of the difference between machine learning and deep learning, this section on the future trends of the two:
1. In view of the increasing use of data science and machine learning in industry, the use of machine learning in business is becoming more and more important for companies that want to survive. At the same time, it is necessary to learn more basic knowledge.
2. Deep learning brings more and more surprises to people, and it will always be like this in the future. Deep learning has proven to be one of the most advanced and best technologies in the existing technology.
3. Deep learning and machine learning and research are still going on, and they are not limited in the academic world as before. At present, machine learning and deep learning are exploding in industry and academia. And receiving more fund support than before, it is likely to become one of the key points of human development.
Your Ads Here

Your Ads Here

Your Ads Here

Your Ads Here

Newer Posts Newer Posts Older Posts Older Posts

Related Posts

Your Ads Here

Loading comments...