Newton's law of cooling
Should AI product managers learn how to use Newton's Law of Cooling to calculate heat?
1. What is Newton’s law of cooling?
Newton's Law of Cooling is an empirical relationship proposed by the British physicist Sir Isaac Newton (1642-1727). It means that the rate of heat loss from an object is proportional to the temperature difference between the object and its surrounding environment. When there is a temperature difference between the surface of an object and its surroundings, the heat dissipated from the unit area per unit time is proportional to the temperature difference, and the proportional coefficient is called the heat transfer coefficient.
The formula is as follows:
in:
-
T(t): the current temperature of the object
-
H: is the surrounding temperature
-
k: is the proportional coefficient
After modification, it becomes:
If H=0, the formula is:
From the above formula, we can see the decay process of Newton's cooling formula. k is the decay coefficient we set ourselves. After t time, the current problem of the object is the product of the initial temperature and the decay rate.
This formula establishes the functional relationship between "temperature" and "time" and easily constructs an "exponential decay" process.
2. Application of Newton’s Cooling Law in AI Scenarios
Newton's cooling law refers to the exponential decay function between "temperature" and "time". In the AI algorithm, the "heat" decay algorithm can be applied according to the Newton's cooling law formula. For example, the cooling algorithm for hot article rankings.
We can imagine "hot article ranking" as a "natural cooling" process:
-
At a certain point in time, all articles in our APP have a current "temperature" (heat value), and the articles are ranked according to the level of "temperature";
-
If users place bounties, likes, reposts, comments, etc. on certain articles, the "temperature" of the article will increase;
-
But as time goes by, it is impossible for us to keep these historical articles at a relatively high "temperature". We need to cool them down so that some new articles can obtain better rankings.
-
We know that we can increase the popularity of articles through bounties, likes, reposts, comments, etc., but we need to find certain methods to reduce the popularity. However, some of the opposite things related to increasing popularity, such as disinterest, reporting, etc., can reduce the popularity. But it is difficult to reduce the popularity based on time, otherwise it will be difficult to rank new articles reasonably during ranking.
Therefore, the decay of article popularity over time is very similar to the decay between temperature and time. We can define that an article has a popularity score,
Current heat score = score of the previous period x exp (- (cooling coefficient) x number of hours between intervals)
The cooling coefficient is defined by us based on business needs. We can assume an initial score, a time interval, a post-time interval score, and calculate the cooling coefficient through these three values. Once the cooling coefficient is calculated, it can be used in the model.
Through the introduction of this law, we can give our product managers an idea. Regarding the requirements related to heat attenuation, we can learn from Newton's cooling law for calculations to ensure the heat and cooling of new and old objects.
Original article: AI product managers should learn how to use Newton’s Law of Cooling to calculate heat?