The inspiration from infant cognitive learning may be the key to the next generation of unsupervised machine learning

Reduce dependency on the collated tag data set , And the desire to take advantage of large amounts of unlabeled data , This has led to unsupervised （ Or self supervision ） New interest in learning algorithms . Although due to the potential representation such as recognition of unraveling 、 The performance is improved by means of comparative learning and clustering optimization , However, unsupervised machine learning still does not reach its hypothetical potential as a breakthrough paradigm to realize general intelligent systems .

cognition （ nerve ） Scientific inspiration is mainly based on adult learners who have access to labels and a large amount of prior knowledge . In order to promote the development of unsupervised machine learning , Trinity University Dublin （Trinity College Dublin） The researchers believe that the developmental science of infant cognition may be the key to the next generation of unsupervised learning methods .

The team identified three key factors that affect the quality and speed of learning in infants ,（1） Infants' information processing is guided and constrained ;（2） Babies are diversifying 、 Learning from multimodal input ;（3） Infant input is shaped by development and active learning .

At the same time, it evaluates the extent to which these insights from infant learning have been used in machine learning , Examined how these implementations are similar to the core insights , And how to further use these factors to improve the previously invisible performance level in unsupervised learning .

The research 「 Lessons from infant learning for unsupervised machine learning」 entitled , On 2022 year 6 month 22 Published on 《 Nature Machine Intelligence》.

Thesis link ： https://www.nature.com/articles/s42256-022-00488-2

Relevant reports ： https://techxplore.com/news/2022-06-babies-key-artificial-intelligence.html