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
Why is machine learning so popular?
Why is machine learning so popular?
Nowadays, the improvement of machine learning is roughly in two aspects. On the one hand, software is the algorithm, from the least squares to the Bayesian thought. On the other hand, hardware, one is parallel computing, such as GPGPU, FPGA; the second is distributed computing, such as Apache's Hadoop, which divides tasks into multiple identical threads and runs applications in large clusters.
Write a picture here
Talking about machine learning always thinks that all are complex mathematical derivations. This is not all right. Big data has both benefits and disadvantages. The benefits are massive, and the harm is redundancy. The massive data you get may be regressed or clustered with just one feature. Often the data is entered into the model before training, and the preparation of the data consumes 80% of the entire process.
Fall in love with data, not learn algorithms.
At the hardware level of the data disposition phase, heterogeneous architectures have been used to stop algorithm acceleration. One is multi-core CPU. The second is dedicated hardware, either using ASIC to stream, or using FPGA to design a wide range of parallel accelerators. At the software level, you use Python or Java, but the interface of the application architecture gives you an easy-to-use framework.
Write a picture here
In statistics, machine learning is nothing more than a statistical application. They ignore the engineering ideas of practice. In the eyes of those who understand business, this is the big idea of feature learning, and neglects mathematical logic.
In machine learning, success is not due to the development of a new algorithm, often in the clever pre-disposition of data, normalization, and the combination of existing methods. As the actual measurement indicates, the effect of using different algorithms is minimal when the data is large enough. This is the idea that data is king. It is also the reason why the data analysts have been hot recently.
Write a picture here
As far as ordinary people are concerned, it is always difficult to learn machine learning before starting. The so-called hardship is always not difficult. From the point of view of data, most of us may rarely become numerators at times, and may even be denominators for life. Here we must talk about the anti-business, that is, dare to bear the blow from time to time, from time to time to bear the impossible, from time to time do not enter and then retreat. In the process, I will gradually recognize myself.
Write a picture here
Talking about machine learning always thinks that all are complex mathematical derivations. This is not all right. Big data has both benefits and disadvantages. The benefits are massive, and the harm is redundancy. The massive data you get may be regressed or clustered with just one feature. Often the data is entered into the model before training, and the preparation of the data consumes 80% of the entire process.
Fall in love with data, not learn algorithms.
At the hardware level of the data disposition phase, heterogeneous architectures have been used to stop algorithm acceleration. One is multi-core CPU. The second is dedicated hardware, either using ASIC to stream, or using FPGA to design a wide range of parallel accelerators. At the software level, you use Python or Java, but the interface of the application architecture gives you an easy-to-use framework.
Write a picture here
In statistics, machine learning is nothing more than a statistical application. They ignore the engineering ideas of practice. In the eyes of those who understand business, this is the big idea of feature learning, and neglects mathematical logic.
In machine learning, success is not due to the development of a new algorithm, often in the clever pre-disposition of data, normalization, and the combination of existing methods. As the actual measurement indicates, the effect of using different algorithms is minimal when the data is large enough. This is the idea that data is king. It is also the reason why the data analysts have been hot recently.
Write a picture here
As far as ordinary people are concerned, it is always difficult to learn machine learning before starting. The so-called hardship is always not difficult. From the point of view of data, most of us may rarely become numerators at times, and may even be denominators for life. Here we must talk about the anti-business, that is, dare to bear the blow from time to time, from time to time to bear the impossible, from time to time do not enter and then retreat. In the process, I will gradually recognize myself.