Google headquarters in Mountain View, California
According to foreign media reports, Google has developed a "multi-functional model that can learn all tasks" and made great progress in the field of AI.
Recently, Google has published a low-key academic paper, which depicts a blueprint for machine learning. Google has developed a new machine learning system called "multi-functional model that can learn all tasks". This model provides a template for future research, how to create a machine learning that can handle multiple tasks well. model.
As Google researchers have said, the "multi-functional model" has undergone a variety of task training, including translation, language analysis, speech recognition, image recognition and target detection. Although its results do not show fundamental improvements to existing methods, at least it shows that training a machine learning system on different tasks helps to improve its overall performance.
Compared to training on a single computing machine, the “multi-function model†has improved accuracy in machine translation, speech and grammar analysis.
Google's papers can provide a case for the development of future machine learning systems that can be applied more widely, and perhaps more accurately than most narrow solutions on the market today. More importantly, these techniques (or their derivatives) can help reduce the amount of training data required to train a viable machine learning algorithm.
The team's research shows that when the “multifunctional model†accepts all the tasks it can do, its accuracy increases as the training data decreases. This is important because it is difficult to accumulate enough training data in some areas.
However, Google does not claim to have an algorithm that can learn all tasks at the same time. As its name implies, the “multifunctional model†is a system tailored to meet different challenges and a system that helps to directly input these expert algorithms. This research does show that the approach taken by Google may be helpful in the future development of the same type of system in different fields.
It's also worth noting that the “multi-function model†has a lot of testing to do. Google's search results have not been confirmed, and it is difficult to know how this research can be extended to other areas. Google's brain team has promised to open up the source code of the "multi-function model", allowing more people to try, but did not give specific opening.
For the "multi-functional model", Google also has some obvious improvements. The Google team pointed out that they did not spend a lot of time optimizing some of the system's fixed parameters (called "hyperparameters" in machine learning), and future tuning optimization will help improve accuracy.
Pharmaceuticals,2-Methyl- Propanoic Acid Monohydrate Price,2-Methyl- Propanoic Acid Monohydrate Free Sample,Pure 2-Methyl- Propanoic Acid Monohydrate
Zhejiang Wild Wind Pharmaceutical Co., Ltd. , https://www.wild-windchem.com