These are very fine ways of explaining simple things in an ego-boosting manner. The more you work with ML these days the more you appreciate it. It happens with every new technology bubble.
In regular terms he's saying the outputs aren't coming out in the same dimensions that the next stages cn work with properly. It wants values between -1 and +1 and it isn't guaranteeing it. Then he's saying you can make it quicker to process by putting the data into a more compact structure for the next stage.
The discriminator could be improved. i.e we could capture better input
KL Diversion is not an accurate tool for manipulating the data, and we have better.
ML is a huge pot of turning regular computer science and maths into intelligible papers. If you'd like assurance, look up something like MinMax functions and Sigmoids. You've likely worked with these since you progressed from HelloWorld.cpp but wouldn't care to shout about them in public
Ahh, people are trying to do a replacement before finding tokens. I wondered why so many people were saying this was difficult.
My head went straight to token parsing, which given the limited set of tokens made it trivial. Thought I was missing something
That article is misinterpreting the meaning of the symbol. It isn't useful in mathematics because it is a contradiction in terms: if "neither of the two compared entities is greater or lesser than the other" then they are equal.
The author of the original article uses it correctly - think about it more in regards to importance for their example.
The business is no more or less important than the developer, but they are NOT equal.
It doesn't have to mean importance though, just the method by which you are comparing things.
Monday ≹ Wednesday
Come to think of it, it should be called the 'No better than' operator.
> if "neither of the two compared entities is greater or lesser than the other" then they are equal.
Not in a partial order.
For example in this simple lattice structure, where lines mark that their top end in greater than their bottom end:
11
/ \
01 10
\ /
00
11 is > to all other (by transitivity for 00), 00 is < to all other (by transitivity for 11), but 01 is not comparable with 10, it is neither lesser nor greater given the described partial order.
You can actually see this kind of structure everyday: unix file permissions for example. Given a user and a file, the permissions of the user are is an element of a lattice where the top element is rwx (or 111 in binary, or 7 in decimal, which means the user has all three permissions to read, write, and execute) and the bottom element is --- (or 000, in binary, or 0 in decimal, which means the user has no permissions). All other combination of r, w, and x are possible, but not always comparable: r-x is not greater nor lesser than rw- in the permissions lattice, it's just different.
Yes, or for more familiar examples: coordinates and complex numbers. the "default" less-than and greater-than don't have any meaning for them; you have to define one, which may be "imperfect" (because one can't do better), hence the concept of partial order.
> It isn't useful in mathematics because it is a contradiction in terms: if "neither of the two compared entities is greater or lesser than the other" then they are equal.
That’s only true for a total order; there are many interesting orders that do not have this property.
It holds for the usual ordering on N, Z, Q and R, but it doesn’t hold for more general partially ordered sets.
In general one has to prove that an order is total, and this is frequently non-trivial: Cantor-Schröder-Bernstein can be seen as a proof that the cardinal numbers have a total order.
Example: alphabetic ordering in most languages with diacritics. For example, "ea" < "éz", but also "éa" < "ez". That's because e and é are treated the same as far as the ordering function is concerned, but they are obviously also NOT the same glyph.
That’s only true for linearly ordered structures, but isn’t true for partially ordered ones.
For example, set inclusion. Two different sets can be neither greater than not smaller than each other. Sets ordered by inclusion form a partially ordered lattice.
The original title on this (which was there when I started this post) was infuriating - but it is still bad.
Is it speaking to the 'true' co-founder who is above the technical one? Maybe the other co-founder who is an 'idea innovator' could up-skill a little and stop pretending that you can build a business by putting on a power suit and making presentations with upturned hands and studied pauses?
You have to rebuild for the latest SDK version; sounds like you have some deprecated code that won't be supported. Stop playing the victim - Apple won't allow apps that won't run to be on the store.
I agree. No issues in exposing your IP. What exactly would be the "fear"? You connect with someone, then they use your IP to file a request with law enforcement to go get your address then use that info to "stalk" you?
That was really fun! would be cool to freeze frame on the blink before disconnect. I couldn't shake the feeling I was just looking at videos until a girl with a nose ring seemed to echo my attempts at not smiling
I disagree about the age. It provides context, and provides inspiration to others. I think he has done a fantastic thing and we don't need to be negative for the sake of being negative
You’re right about the inspirational aspect, but when highlighted by the author themselves it does feel a bit unnecessary, and it detracts from the project. Half of the comments are on the subject of age.
Off-topic for this thread but if you can get the ball to ping-pong horizontally while both players are at the top edge then you can stay alive indefinitely - the opponent doesn't seem to move downwards out of this position.
The fact that the string "12345678" occurs only once in the first 200M digits already gives an impression of the amount of information you may find/store there.
In regular terms he's saying the outputs aren't coming out in the same dimensions that the next stages cn work with properly. It wants values between -1 and +1 and it isn't guaranteeing it. Then he's saying you can make it quicker to process by putting the data into a more compact structure for the next stage.
The discriminator could be improved. i.e we could capture better input
KL Diversion is not an accurate tool for manipulating the data, and we have better.
ML is a huge pot of turning regular computer science and maths into intelligible papers. If you'd like assurance, look up something like MinMax functions and Sigmoids. You've likely worked with these since you progressed from HelloWorld.cpp but wouldn't care to shout about them in public