When we have good, wonderful days in our relationships with our family and friends, we feel happy and look for ways to show that we are satisfied. But when we face problems or encounter some difficulty, do we cut off contact and end the relationship?
First of all, let's consider that human emotions and feelings, like other information, occupy a certain part of our memory. While some of them (excitement, surprise, anxiety, confusion, etc.) temporarily occupy a place in our RAM, some (love, remorse, resentment, trust, etc.) are written to our ROM memory and, no matter how much time passes, they sit in a corner waiting to be used. No matter how much we are not aware of it, this affects the decisions we make and the reactions we give.
How? Very simple: With a function. This function (hereinafter referred to as a model) plays the role of a map in our relationships.
In Machine Learning, we train and test our data, sometimes multiple times, by manipulating our data with certain weight(w) and bias(b) values to find the best fit model. While positive emotions in the encoded data help improve the accuracy score of our model, negative emotions keep us away from the problem of over-fitting. Over-fitting means that even though our model learns our training data well, it does not produce correct results in our test data. Vice versa, if we do not properly incorporate our positive emotions into our data, and negative emotions become overwhelming, it will lead to the problem of underfitting (training a model that is too simple or cannot learn at all).
It is the same in relationships: the emotions or feelings from past events are the inputs (x) in our training data, and our reactions are our outputs (y or f(x)). When we encounter a new event, we put the inputs (emotions and feelings) into a model function that is trained on our training data in our brain, and test it. The result is our reaction to the new event. In fact, without us realizing it, the inputs and outputs in this new event are transformed into our training data from time to time, helping to improve our model, and in each subsequent event, in each subsequent situation, we call that model, query it, make adjustments, and send it back.
Let's look at the same example situations:
· Let's imagine that a young man has been working in a company for several years. The company gives corporate gifts to its employees on every holiday and organizes various entertainment or incentive events. Another holiday is coming, and like every employee, this young man already has an expectation from the company on the occasion of the holiday. Since the company's financial situation is not good, this year, in order to reduce additional expenses, congratulatory messages are sent only via corporate email.
What happened? Since our model was overfit in our training data, the output we received when entering the input in our test data did not coincide with the real result. That is, the employee was mistaken about congratulating him on this holiday, and our model got the wrong result.
· If the company had never given corporate gifts or organized various events, then the employee would have no expectations for the holiday. If the company had made a difference and organized something for this holiday, the model would still be considered to have received the wrong result because it did not coincide with the employee's expectation. In this case, our model gets the wrong result because it is under-fitting.
Sometimes, no matter how good our model is and how successful our tests are, we see that our model has difficulty showing the correct result in some situations. There are some emotions that we do not always encounter, or they are definitely the first of something. What happens when this happens? Our data is considered an outlier. For example, a person loses a family member once, and this does not happen repeatedly, or the feeling of becoming a mother or father for the first time is never repeated, and a person does not experience the moment of taking off for the first time like the first time, neither before nor after. We usually do not take such data into account in every reaction, in every decision. Why usually? Because there is no such thing as “all outliers should be discarded.” We try to avoid risks by considering them. For example, when someone has a heart attack, it is already a signal to him and those around him that tension and stress levels should be reduced, attention should be paid to diet, and regular exercise should be done. We were talking about relationships: if some of your behavior made a friend distance himself from you, then something is wrong, and you need to fix it.
So what is ideal? Working with balanced data. We try to derive the right model, the formula for our relationships, taking into account outliers when necessary, sacrificing them when necessary, starting with underfitting at the beginning and going to overfitting in the middle. By making adjustments in each new situation, we get a better model than the previous one. That is why we laugh at our previous decisions, admitting that they were childish, and making sentences like “If I had the mind I have now, I would have done this or that.” That is, if we had the current model, we would have made the previous decisions completely differently. But wasn’t this model created by their manipulation? If we changed them, wouldn’t our model be different?
So how do you manipulate values to build your own model? What did you prioritize and sacrifice to get the best model?
Author: Gunel Khosrovlu
First of all, let's consider that human emotions and feelings, like other information, occupy a certain part of our memory. While some of them (excitement, surprise, anxiety, confusion, etc.) temporarily occupy a place in our RAM, some (love, remorse, resentment, trust, etc.) are written to our ROM memory and, no matter how much time passes, they sit in a corner waiting to be used. No matter how much we are not aware of it, this affects the decisions we make and the reactions we give.
How? Very simple: With a function. This function (hereinafter referred to as a model) plays the role of a map in our relationships.
In Machine Learning, we train and test our data, sometimes multiple times, by manipulating our data with certain weight(w) and bias(b) values to find the best fit model. While positive emotions in the encoded data help improve the accuracy score of our model, negative emotions keep us away from the problem of over-fitting. Over-fitting means that even though our model learns our training data well, it does not produce correct results in our test data. Vice versa, if we do not properly incorporate our positive emotions into our data, and negative emotions become overwhelming, it will lead to the problem of underfitting (training a model that is too simple or cannot learn at all).
It is the same in relationships: the emotions or feelings from past events are the inputs (x) in our training data, and our reactions are our outputs (y or f(x)). When we encounter a new event, we put the inputs (emotions and feelings) into a model function that is trained on our training data in our brain, and test it. The result is our reaction to the new event. In fact, without us realizing it, the inputs and outputs in this new event are transformed into our training data from time to time, helping to improve our model, and in each subsequent event, in each subsequent situation, we call that model, query it, make adjustments, and send it back.
Let's look at the same example situations:
· Let's imagine that a young man has been working in a company for several years. The company gives corporate gifts to its employees on every holiday and organizes various entertainment or incentive events. Another holiday is coming, and like every employee, this young man already has an expectation from the company on the occasion of the holiday. Since the company's financial situation is not good, this year, in order to reduce additional expenses, congratulatory messages are sent only via corporate email.
What happened? Since our model was overfit in our training data, the output we received when entering the input in our test data did not coincide with the real result. That is, the employee was mistaken about congratulating him on this holiday, and our model got the wrong result.
· If the company had never given corporate gifts or organized various events, then the employee would have no expectations for the holiday. If the company had made a difference and organized something for this holiday, the model would still be considered to have received the wrong result because it did not coincide with the employee's expectation. In this case, our model gets the wrong result because it is under-fitting.
Sometimes, no matter how good our model is and how successful our tests are, we see that our model has difficulty showing the correct result in some situations. There are some emotions that we do not always encounter, or they are definitely the first of something. What happens when this happens? Our data is considered an outlier. For example, a person loses a family member once, and this does not happen repeatedly, or the feeling of becoming a mother or father for the first time is never repeated, and a person does not experience the moment of taking off for the first time like the first time, neither before nor after. We usually do not take such data into account in every reaction, in every decision. Why usually? Because there is no such thing as “all outliers should be discarded.” We try to avoid risks by considering them. For example, when someone has a heart attack, it is already a signal to him and those around him that tension and stress levels should be reduced, attention should be paid to diet, and regular exercise should be done. We were talking about relationships: if some of your behavior made a friend distance himself from you, then something is wrong, and you need to fix it.
So what is ideal? Working with balanced data. We try to derive the right model, the formula for our relationships, taking into account outliers when necessary, sacrificing them when necessary, starting with underfitting at the beginning and going to overfitting in the middle. By making adjustments in each new situation, we get a better model than the previous one. That is why we laugh at our previous decisions, admitting that they were childish, and making sentences like “If I had the mind I have now, I would have done this or that.” That is, if we had the current model, we would have made the previous decisions completely differently. But wasn’t this model created by their manipulation? If we changed them, wouldn’t our model be different?
So how do you manipulate values to build your own model? What did you prioritize and sacrifice to get the best model?
Author: Gunel Khosrovlu