FPGA Machine Learning

After two months of understanding about machine learning, I found that machine learning has a variety of directions. Webpage sorting, speech recognition, image recognition, and recommendation system. There are also a variety of algorithms. After reading other books, I found that in addition to K-means clustering, Bayesian, neural networks, online learning, and so on, there are many other algorithms. For example, immune algorithms, genetic algorithms, principal component analysis, and ant colony algorithms. It seems that many algorithms require a lot of research before they can be used very well. It is said that deep learning is upgraded by neural networks. Neural networks are a book with a lot of content. The Dragon Star program also involves the application of multiple algorithms. Do you want to follow popular algorithms or find the latest machine learning algorithms ??

Deep Learning is a hot topic recently. There are more materials and more people to learn. Or is it a rare immune algorithm, ant colony algorithm ??? From a performance perspective, deep learning has good performance, but immune algorithms may improve performance in the next two years. In this case, what is better ?? In my opinion, if you have profound mathematical skills, good thinking, and many creative friends, I suggest you develop new algorithms, like immune algorithms. Of course, it would be better to create a bee nest algorithm. It is estimated that many people do not have this condition, so we can become a follower. Select the popular deep learning algorithms. It should be good to find a scenario and company for deep learning applications.

I may still feel a little different about the ai I can do. I don't want to say that robots beat humans, and there are a lot of robots in sci-fi movies. I don't have that skill. What I want to do is very simple. Let the machine's eyes understand common things and do some simple things. So my main direction is machine vision. So how do I plan to move forward step by step? Or what do I want to learn?

My current content is about image processing. In fact, image processing is the front-end processing of pattern recognition, so that the image features can be better reflected. The next step is pattern recognition, which can only be understood in a narrow sense. It is feature extraction. In fact, it has entered the scope of machine learning. Finally, machine learning can be integrated into cognition. A lot of FPGA processing chips are designed here (this will be said later ). From another angle, I want to learn the content, image processing, such as contrast, Image Correction, and border scan. Machine Learning is suitable for image applications from a variety of learning algorithms, such as deep learning and principal component analysis. Applications and simple algorithms ). Machine Learning can also achieve image processing content, for example, clustering can be used for image segmentation. But why do we need to learn image processing technologies from time to time ?? The idea is that machine learning is the process of Automatically Extracting Features. For example, you may know the classification process and feature extraction process of a decision tree. However, you often do not know the process, however, image processing is artificially supplied, separated, and has some special features. This may reduce the difficulty of machine learning (purely conjecture, and lack of knowledge about machine learning ).

What about FPGA ??? The main consideration is the computing speed. At present, FPGA computing speed is the best, such as: drone rescue in the disaster area, flight speed, camera pixels, recognition, it requires a lot of information about the computing positioning personnel. For example, the training time, speed is an important indicator. However, FPGA cannot complete complex computing. If the GPU or APU's computing power can be upgraded to another level on that day, I will also consider learning.

These are only the main learning content, and there are many small things to keep up with, such as mathematics. There is a lot of content. I can only compare my skills in the time I can grasp to balance the learning time of each part. These are what I want to say, and my friends who want to learn with me. Let's study with me. My QQ, 849886241. Please pay attention and ask for help. The road is long. Ask for help.