News classification
Contact us
- Add: No. 9, North Fourth Ring Road, Haidian District, Beijing. It mainly includes face recognition, living detection, ID card recognition, bank card recognition, business card recognition, license plate recognition, OCR recognition, and intelligent recognition technology.
- Tel: 13146317170 廖经理
- Fax:
- Email: 398017534@qq.com
What is the machine learning in AI?
What is the machine learning in AI?
What is machine learning?
We will find that there are many experience based prejudgments. For example, why do you see a slight temperature, a breeze, a sunset, think that tomorrow is a good day? This is due to a lot of similar conditions in our life experience. After the first day to see the above features, the weather is usually very good for the second day. OK.
Why is the color green, the root curling, the sound of the sound, we can distinguish a good melon? Because we have eaten and read a lot of watermelons, so we can make quite good discrimination based on the features of color, roots and sound. With the concept and homework done, we will get good results.
Can we see that we can make an effective decision? Is that we have accumulated a lot of experience, and we can make effective decisions on the new situation through the application of the experience. The application of the experience is done by our human itself. Can the computer help?
Machine learning is a subject that tries to discuss how to pass the calculated wrist and apply experience to improve the performance of the system itself in a computer system. "Experience" usually exists in a "data" way. Therefore, the main content of the machine learning is to generate "model" from the data on the computer (MOD EL) algorithm, namely learning algorithm. With the learning algorithm, we provide it with the experience data, and it can generate models based on these data; in the face of new conditions (such as seeing a watermelon that is not open), the model will provide us with the corresponding discrimination (such as good melon). If computer science is to discuss the knowledge of algorithm, then it can be said that machine learning is the study of "learning algorithm".
Fundamental terminology
To stop learning from machines, you have to have data first. Suppose we have collected a batch of data on watermelons, such as green color = green; root = curling; sound = turbid), (color = pitch black; root = slightly curled; sound = dull) (tone = shallow; root = stiff; knocking = hung),... Each pair of brackets is a record, "=" meaning is "value".
The confluence of this set of records is called a "data set", each of which is a description of a thing or object (a watermelon), known as a "instance" or "Samp1e". A matter that reflects the performance or nature of a thing or object in a certain aspect, such as "color", "root" "knocking", called "attribute" or "feature"; attribute values, such as "green" and "black", are called "attribute va1ue". The space of attribute Chang Cheng is called "attribute space", "Samp1e space" or "input space". For example, we take the "color" "root" "knocking" as the three axis, and they form a three-dimensional space to describe the watermelon, each watermelon can find the position of their own coordinates in this space. Since every point in space corresponds to a coordinate vector, we also refer to an example as a "feature vector".
In general, D = {X1, X2 ".., Xm}" represents a set of data containing m examples, each of which is depicted by D properties (for example, the watermelon data above uses 3 attributes), and each example Xi = (Xi1; Xi2;.; Xid) is a vector in D dimension sample space X. The values of attributes (for example, the above third watermelon values on second attributes are "stiff"), and D is called the "dimension" (dimensionality) of sample Xi.
The process of learning the data from data is called learning or training, which is implemented by implementing a learning algorithm. The data used during the exercise process is called "training data", in which each sample is called an exercise sample (training Samp1e), and the confluence of the exercise sample is called "training set".
The learning model corresponds to some potential law about data, so it is also called "hypothesis"; the potential law itself is called "truth" or "truth" (ground-truth). The learning process is to find out or close to the truth, and can be regarded as the instantiation of the learning method in the given data and parameter space.
If you want to learn a model that can help us to identify not a "good melon", only the previous example data is obviously not enough to set up such a prediction model, we need to get the "result" information of the exercise sample, such as "(color = green; root = curling; knocking = turbid), Good melon. The information on the results of the example, such as "good melon", is called "labe1"; an example with a flag information is called a "examp1e". Generally, the I sample is represented with (Xi, Yi). Among them, Yi e Y is the symbol of the example Xi, Y is the confluence of all signs, also known as "logo space" (label space) or "output space".
If we want to predict discrete values, such as "good melon" and "bad melon", this kind of learning task is called "classification".
If you want to predict the continuous value, such as watermelon maturity 0.95 and 0.37, such learning task is called "regression".
For the "two classification" (binary classification) task that only touches two categories, one of them is usually called the "positive class" and the other is the "negative class"; when it touches multiple categories, it is called
We will find that there are many experience based prejudgments. For example, why do you see a slight temperature, a breeze, a sunset, think that tomorrow is a good day? This is due to a lot of similar conditions in our life experience. After the first day to see the above features, the weather is usually very good for the second day. OK.
Why is the color green, the root curling, the sound of the sound, we can distinguish a good melon? Because we have eaten and read a lot of watermelons, so we can make quite good discrimination based on the features of color, roots and sound. With the concept and homework done, we will get good results.
Can we see that we can make an effective decision? Is that we have accumulated a lot of experience, and we can make effective decisions on the new situation through the application of the experience. The application of the experience is done by our human itself. Can the computer help?
Machine learning is a subject that tries to discuss how to pass the calculated wrist and apply experience to improve the performance of the system itself in a computer system. "Experience" usually exists in a "data" way. Therefore, the main content of the machine learning is to generate "model" from the data on the computer (MOD EL) algorithm, namely learning algorithm. With the learning algorithm, we provide it with the experience data, and it can generate models based on these data; in the face of new conditions (such as seeing a watermelon that is not open), the model will provide us with the corresponding discrimination (such as good melon). If computer science is to discuss the knowledge of algorithm, then it can be said that machine learning is the study of "learning algorithm".
Fundamental terminology
To stop learning from machines, you have to have data first. Suppose we have collected a batch of data on watermelons, such as green color = green; root = curling; sound = turbid), (color = pitch black; root = slightly curled; sound = dull) (tone = shallow; root = stiff; knocking = hung),... Each pair of brackets is a record, "=" meaning is "value".
The confluence of this set of records is called a "data set", each of which is a description of a thing or object (a watermelon), known as a "instance" or "Samp1e". A matter that reflects the performance or nature of a thing or object in a certain aspect, such as "color", "root" "knocking", called "attribute" or "feature"; attribute values, such as "green" and "black", are called "attribute va1ue". The space of attribute Chang Cheng is called "attribute space", "Samp1e space" or "input space". For example, we take the "color" "root" "knocking" as the three axis, and they form a three-dimensional space to describe the watermelon, each watermelon can find the position of their own coordinates in this space. Since every point in space corresponds to a coordinate vector, we also refer to an example as a "feature vector".
In general, D = {X1, X2 ".., Xm}" represents a set of data containing m examples, each of which is depicted by D properties (for example, the watermelon data above uses 3 attributes), and each example Xi = (Xi1; Xi2;.; Xid) is a vector in D dimension sample space X. The values of attributes (for example, the above third watermelon values on second attributes are "stiff"), and D is called the "dimension" (dimensionality) of sample Xi.
The process of learning the data from data is called learning or training, which is implemented by implementing a learning algorithm. The data used during the exercise process is called "training data", in which each sample is called an exercise sample (training Samp1e), and the confluence of the exercise sample is called "training set".
The learning model corresponds to some potential law about data, so it is also called "hypothesis"; the potential law itself is called "truth" or "truth" (ground-truth). The learning process is to find out or close to the truth, and can be regarded as the instantiation of the learning method in the given data and parameter space.
If you want to learn a model that can help us to identify not a "good melon", only the previous example data is obviously not enough to set up such a prediction model, we need to get the "result" information of the exercise sample, such as "(color = green; root = curling; knocking = turbid), Good melon. The information on the results of the example, such as "good melon", is called "labe1"; an example with a flag information is called a "examp1e". Generally, the I sample is represented with (Xi, Yi). Among them, Yi e Y is the symbol of the example Xi, Y is the confluence of all signs, also known as "logo space" (label space) or "output space".
If we want to predict discrete values, such as "good melon" and "bad melon", this kind of learning task is called "classification".
If you want to predict the continuous value, such as watermelon maturity 0.95 and 0.37, such learning task is called "regression".
For the "two classification" (binary classification) task that only touches two categories, one of them is usually called the "positive class" and the other is the "negative class"; when it touches multiple categories, it is called