Google quietly released an academic paper that could provide a blueprint for the future of machine learning. Called One Model to Learn Them All, it lays out a template for how to create a single machine learning model that can address multiple tasks well. <\/p>\n
The MultiModel, as the Google researchers call it, was trained on a variety of tasks, including translation, language parsing, speech recognition, image recognition, and object detection. While its results dont show radical improvements over existing approaches, they illustrate that training a machine learning system on a variety of tasks could help boost its overall performance. <\/p>\n
For example, the MultiModel improved its accuracy on machine translation, speech, and parsing tasks when trained on all of the operations it was capable of, compared to when the model was just trained on one operation. <\/p>\n
Googles paper could provide a template for the development of future machine learning systems that are more broadly applicable, and potentially more accurate, than the narrow solutions that populate much of the market today. Whats more, these techniques (or those they spawn) could help reduce the amount of training data needed to create a viable machine learning algorithm. <\/p>\n
Thats because the teams results show that when the MultiModel is trained on all the tasks its capable of, its accuracy improves on tasks with less training data. Thats important, since it can be difficult to accumulate a sizable enough set of training data in some domains. <\/p>\n
However, Google doesnt claim to havea master algorithm that can learn everything at once. As its name implies, the MultiModel network includes systems that are tailor-made to address different challenges, along with systems that help direct input to those expert algorithms. This research does show that the approach Google took could be useful for future development of similar systems that address different domains. <\/p>\n
Its also worth noting that theres plenty more testing to be done. Googles results havent been verified, and its hard to know how well this research generalizes to other fields. The Google Brain team has released the MultiModel code as part of the TensorFlow open source project, so other people can experiment with it and find out. <\/p>\n
Google also has some clear paths to improvement. The team pointed out that they didnt spend a lot of time optimizing some of the systems fixed parameters (known as hyperparameters in machine learning speak), and going through more extensive tweaking could help improve accuracy in the future. <\/p>\n
Updated 10:45: This story initially said that there was not a timetable for releasing the MultiModel code under an open source license. The code was released last week. This story has been updated to note that and include a link to the repository. <\/p>\n
<\/p>\n
Here is the original post: <\/p>\n
Google advances AI with 'one model to learn them all' - VentureBeat<\/a><\/p>\n","protected":false},"excerpt":{"rendered":" Google quietly released an academic paper that could provide a blueprint for the future of machine learning. Called One Model to Learn Them All, it lays out a template for how to create a single machine learning model that can address multiple tasks well. The MultiModel, as the Google researchers call it, was trained on a variety of tasks, including translation, language parsing, speech recognition, image recognition, and object detection. Continue reading