https://mp.weixin.qq.com/s/UKEnUAX-Kwiv3kmjVeYHPw
By 超神经
场景描述：蜜蜂对作物的生产影响巨大，但在近些年，蜜蜂的种群数量却在不断减少，面对这种情况，机器学习、数据分析等方法的介入，能够帮助蜜蜂避免走向灭绝的道路。
关键词：物种保护 机器学习 数据分析 There was once a saying: "After the bees are extinct, humans can only live for four years." This sentence was once believed to be from Einstein.
Did Einstein really say it?
Why is there such a statement? Of course not because there is no honey to eat. The meaning of bees to nature and humans is far more than you think.
In addition to bringing delicious and nutritious honey, the main contribution of bees is the spread of pollen. Pollinators such as bees, birds and bats affect 35% of global crop production and help 87% of the world’s major food crops and many plant drugs to increase production.
Regarding the statement that "bees are extinct, humans will not survive for four years", although it has a warning effect, it does not come from Einstein and its accuracy has also been falsified.
But as an important part of nature, if bees are missing, it will trigger a domino-like effect in the food chain. Insect-borne crops will be severely affected, and people may face food shortages and other problems.
In recent years, the population survival of bees is facing a severe test.
According to a report by the World Bee Project (WBP), the number of bee colonies in the world has declined sharply in recent years, and bees and other pollinating insects are on the verge of extinction.
Fortunately, with the development of science and technology, the application of machine learning and other technologies has brought hope for bee protection.
Why did the number of hardworking bees plummet?
A large part of crop cultivation is inseparable from the credit of bees. For crops such as apples, broccoli, blueberries, and onions, 90% rely on bees.
In addition, the spread of bees is 80% of cherries and celery, 65% of watermelons and plums, and 45% of tangerines. Even lemons and cotton rely on bees for 20%.
A bee must collect 4 million flowers and fly a distance equivalent to four times around the earth to collect one kilogram of honey
But beginning in 2006, the United States first noticed that the number of bees had plummeted. Adult bees abandoned the hive and then flew to death. Subsequently, Belgium, France, Germany, the Netherlands and other countries also discovered this phenomenon.
This phenomenon of the disappearance of the worker bees in a large number of hive and the ecological collapse of the colony was finally named "Colony Collapse Disorder" (CCD for short).
According to data from the German Beekeepers Association in 2018, more than 300 of the 560 species of bees in Germany are at risk of extinction, and the number of bee colonies has fallen from 2.5 million colonies in 1952 to less than 1 million colonies today.
However, the reason for the sudden decrease in the number of bees has not been fully understood. It may come from various factors, such as the threat of foreign insects, the abuse of pesticides, the deterioration of the natural environment, and the disadvantages caused by single planting.
In order to curb this phenomenon, bee protection has gradually been put on the agenda. In December 2017, the United Nations designated May 20 every year as World Bee Day.
Can robot bees replace bees?
The reduction in the number of bees has the most direct impact on crop yields. In some reports, some orchards face the problem of blocked pollination, and they have no choice but to use a large number of workers to act as "bee people," climbing up fruit trees and manually pollinating with brushes and other tools.
Orchard workers pollinate trees
There are also some researchers who have come up with the idea of "mechanical bees" and want to use machines to take on the work of bees.
For example, a technician named Toeijiro in Japan manufactured a mechanical pollinator drone in 2017.
It is a 4 cm square, 15 gram mini-aircraft with horse hair glued to the bottom to simulate the hairy trunk of a bee. A layer of special viscous gel is applied to the horse hair to help the spread of pollen.
Toeijiro's pollen drone pollination display
This unmanned aerial vehicle achieves pollination operation, but the disadvantage is that it requires manual remote control, and considering economy and applicability, it cannot be widely promoted.
In 2018, Dutch scientists did similar work and invented a bee robot called DelFly, which has greatly improved flying technology, but it is still limited to use in greenhouses.
DelFly robot appearance
Although these researchers have made interesting attempts, the idea of circumventing bees as a part of the natural world is still unrealistic.
AI program guarding bees
Perhaps the most reliable way is to return to the protection of bees. In this regard, machine learning can play its advantages.
An application called Bee Health Guru can be accessed normally on mobile phones. It is based on machine learning algorithms and can help beekeepers understand the health of the hive.
Various abnormal information displayed by Bee Health Guru
The main principle of this program is to judge the abnormality of the bee colony based on the sound information. If there is a problem with the honeycomb, it will produce a special sound. Use the program to "listen" to find the cause of the problem.
In order to make the model more accurate, they spent five years collecting a lot of sounds. In use, by recording the sound, distinguish the nuances and provide timely warning to the beekeeper.
Bee Health Guru APP interface
Compared with traditional workers' judgment, Bee Health Guru can track the situation of multiple variables, provide beekeepers with health information of the bee colony, and can also stifle some unfavorable factors in the budding state to stabilize the condition of the beehive.
In practice, this approach has proved to be very effective, for example, it can avoid the accidental death of queen bees and detect the spread of viruses in the colony in time.
Create an intelligent honeycomb system
In addition, some research teams are studying and collecting other indicators to reveal the health change data of the hive, including brooding temperature, environmental temperature, hive weight, humidity and group behavior. Through this information, open the way to protect bees.
Among them, Oracle Corporation and the University of Reading in the United Kingdom developed a honeycomb monitoring system in a collaborative study. The system mainly includes six smart sensors that can collect a large number of different information about the honeycomb.
The sensor will provide data on the sound, temperature, humidity and weight of the honey in the honeycomb
Researchers use key information, such as wing movement and motion footprints, to combine other measurements, such as temperature, humidity, and honey production, and then use machine learning algorithms to predict modeling based on historical events.
In this way, they built an intelligent management system. The system can closely monitor the colony, develop effective detection modes and predict behavior. In addition, it can judge the impact of certain events on the bee population and send warnings of potential threats to beekeepers.
Last fall, Oracle deployed this system in honeycombs in the UK, Australia and Israel.
Save the bees, not for honey
Last year, May 20th was the first World Bee Day, but many people still don't know the existence of this holiday. Perhaps the extinction of bees is still far away for many people, but the consequences of species extinction have been described in detail in the book "Silent Spring".
The British poet Alexander Pope wrote, "No matter which link you hit in the chain of nature, the tenth or tenth thousandth will break it." While protecting the bees, we are actually protecting ourselves. .
The use of science and technology allows us to gain more insights from data, bring good suggestions for bee breeding, and grasp the gathering state of bees in real time, so as to help bees develop well and create a sustainable ecology. system.
In many places, technology such as machine learning is only used to seek the well-being of human beings, but the nature on which we live has been ignored.
The concern of technology for the issue of bee extinction may be returning to its roots. By helping species to multiply, it not only allows us to embrace a more complete nature, but also allows us to feel that human beings are not selfish and lonely in the warmth of technology. .
HyperNeuropedia
Bi-directional long-short-term memory
Bi-directional Long-Short Term Memory
Bi-directional long short-term memory (Bi-LSTM) is an extension of RNN.
The hidden layer of the bidirectional convolutional neural network needs to store two values, A participates in the forward calculation, and A'participates in the reverse calculation.
The final output value y depends on these two values.
Long-term memory (LSTM) is to replace the module of the hidden layer in the recurrent neural network with the module of long- and short-term memory.
The two-way LSTM is superior to the one-way LSTM in that it can use the information in both the past and future moments at the same time, so that the final prediction is more accurate.
