Leveraging knowledge of the gang in ML fashions.
Lately, we take sources reminiscent of Wikipedia or Reddit with no consideration — these sources depend on the collective data of particular person contributors to serve us with largely correct info, which is usually known as the “knowledge of the gang”. The concept is that the collective choice will be extra correct than any particular person’s judgement, since we are able to every have our personal implicit biases and/or lack of awareness, leading to some degree of error in our judgement. Collectively, these errors is perhaps offset by one another — for instance, we are able to compensate for another person’s lack of awareness/experience in a single space, whereas they’d make up for ours in different areas. Making use of this concept to machine studying leads to “ensemble” strategies.
At a really excessive degree, we practice machine studying fashions with a purpose to make predictions in regards to the future. In different phrases, we offer the fashions with the coaching knowledge with the hope that mannequin could make good predictions in regards to the future. However what if we may practice a number of machine studying fashions after which one way or the other mixture their opinions in regards to the predictions? It seems, this generally is a very helpful method, which is broadly used within the trade.