Well, let me tell you about this hardmax thing. It’s one of them new-fangled computer things, I reckon. Heard some young folks talking about it the other day. Sounds complicated, but I’ll try to explain it like I see it.
So, this hardmax, it’s like picking the biggest apple from the basket. You got a whole bunch of apples, different sizes. You just pick the biggest one, right? That’s what this hardmax does, but with numbers.
Imagine you got a bunch of numbers, like, say, 2, 5, 1, and 8. This hardmax thing, it looks at all them numbers and picks out the biggest one, which is 8 in this case. Then, it changes that 8 to a 1, and all the other numbers, it makes ’em 0. So you end up with 0, 0, 0, and 1. Simple, ain’t it?
Now, there’s another thing called softmax. This one is a bit more tricky. It’s like, instead of just picking the biggest apple, you’re kinda weighing all the apples. You give the biggest apple the most weight, but the other apples still get some weight, too. It’s like saying, “This big apple is the most likely, but these other ones, they got a chance too.”
With numbers, softmax makes ’em all add up to 1, like a whole pie. Each number gets a slice of the pie, and the bigger the number, the bigger the slice. It’s all about making things into probabilities, these young folks say. Whatever that means. It’s just like, this number is more likely than that one.
- Hardmax picks the biggest number.
- Softmax gives all numbers a chance.
- Hardmax makes the biggest number 1, others 0.
- Softmax makes all numbers add up to 1.
This hardmax is like a shortcut, I guess. It’s quicker, just pick the biggest and be done with it. Softmax, it’s more like thinking about all the possibilities, even the small ones. It’s like when you have to decide between going left or right and you consider all pros and cons before making a decision. This is softmax. It is more difficult than hardmax, but sometimes we have to use it.
They say this hardmax is used in something called a “one-hot encoding.” Sounds like something to do with a poker game, but I don’t know. It’s just a way of saying, “This one is the winner, and the rest ain’t.” Like picking the winning lottery number. Only one number wins, the rest are just numbers. So when you want to use hardmax, you can do it.
Now, why would you use one over the other? Well, I reckon it depends on what you’re trying to do. If you just need to pick the biggest one, no fuss, no muss, then hardmax is your thing. It’s like picking the best horse in a race. You just want to know who won, right?
But if you need to be a bit more careful, consider all the options, then softmax is the way to go. It’s like deciding what to cook for supper. You might have a favorite, but you gotta think about what everyone else likes, too.
This whole thing is used in something called “deep learning.” Now that’s a real head-scratcher. It’s like teaching a computer to think, I guess. They show it a bunch of stuff, and it learns to recognize things. Like showing a dog a bunch of pictures of cats, and eventually, it learns to tell a cat from a dog. Hardmax and softmax are just some of the tools they use to do this.
They also talk about something called a “step activation function”. It’s like a light switch. It’s either on or off, 1 or 0. If the number is big enough, the light turns on. If not, it stays off. Seems simple enough, even for an old gal like me.
There is another thing called “hard limit function”. This one can make things become 1 or 0. If the number is big enough, it will become 1, or it will be 0. This one is very useful when we use computer to do calculation.
These young folks, they come up with all sorts of fancy names for these things. But when you break it down, it ain’t so complicated. It’s just different ways of looking at numbers and making decisions. Like life, I guess. Sometimes you gotta go with your gut, pick the biggest apple. Other times, you gotta weigh all your options, give everything a fair shake. Hardmax is used in many places. That is what I know about it.
