The whole series of the lectures are so great and I like them very very much! As you mentioned in one of your lecture, there are three big theorems in information theory: source coding theorem, rate distortion theorem and channel coding theorem. It is my best wish that you can provide video lectures for the contents of the channel coding and rate distortion theorems. Thank you so much for the wonderful lectures!
Hello, I have watched the whole series of lectures about arithmetic encoding and I'm very thrilled about this algorithm and your way of explaining things. You are a great professor and mathematician. Thank you for your lectures! Unfortunately my finite precision implementation works incorrectly on large input data (maybe due to round off error). So maybe you cold share your implementation of the algorithm. It would be a great help! Thank you in advance!
Set the probability distribution so that each p(x_i) has the form 1/2**m with m>=0. It seems the algorithm work for long sequence in this case. I guess this can get rid of the round off error issue but I haven't proved it yet. In addition, any places like "< half" or ">half" or ">quarter and
Hi, Professor Miller! Could you teach me the Rate-Distortion theory? I'm a ph.D. student who wishes to combine it into the field of deep learning. but my major is computer science, and it was very hard for me to grab the information theory. I found your video lectures on the information theory is the most understandable material for me. I hope you could take some time to record the lectures on Rate-Distortion Theory. I found your lecture note about it, but could not understand it without your explanation.
@Cube Earth I know this is old and you probably figured it out but he was basically the copying the first precision number of bits from the encoded message into z.
If precision is say 32, then whole is 2^32, when finding an approximation of z, if in the first iteration beta(i)=1, then Z = 2^32.... which means for any other 1 in the sequence Z will be more than whole. Am I misunderstanding? And if not, how can Z be in the range of [a,b)?
Hello, once I watched your intro to logistic regression (which was so great) but I cannot find it anymore, would you provide me with a link? Thanks so much!
*The pseudocode presented in this video contains one mistake that makes any implementation from it useless* Here is my implementation of this algorithm in C++. github.com/KyrietS/arithmetic-coding/blob/f80e7a911e898a45d114a1103149ad3b5feb0f40/AC_Core/src/AdaptiveScalingCoder.cpp The structure is the same as presented in this video. Look at the line '123' in my source code. THERE MUST BE A "BREAK" KEYWORD. I spent hours of debugging because of it... I've also improved this example and made it more efficient using Binary Search for decoding symbols. github.com/KyrietS/arithmetic-coding/blob/1584592fa666c55b3fd82b5af99afa5d9089743c/AC_Core/src/AdaptiveScalingCoder.cpp You can find my newest version of Adaptive Scaling Arithmetic Coding on my master branch: github.com/KyrietS/arithmetic-coding
The whole series of the lectures are so great and I like them very very much! As you mentioned in one of your lecture, there are three big theorems in information theory: source coding theorem, rate distortion theorem and channel coding theorem. It is my best wish that you can provide video lectures for the contents of the channel coding and rate distortion theorems. Thank you so much for the wonderful lectures!
Hello, I have watched the whole series of lectures about arithmetic encoding and I'm very thrilled about this algorithm and your way of explaining things. You are a great professor and mathematician. Thank you for your lectures! Unfortunately my finite precision implementation works incorrectly on large input data (maybe due to round off error). So maybe you cold share your implementation of the algorithm. It would be a great help! Thank you in advance!
Set the probability distribution so that each p(x_i) has the form 1/2**m with m>=0. It seems the algorithm work for long sequence in this case. I guess this can get rid of the round off error issue but I haven't proved it yet. In addition, any places like "< half" or ">half" or ">quarter and
Hi, Professor Miller! Could you teach me the Rate-Distortion theory?
I'm a ph.D. student who wishes to combine it into the field of deep learning.
but my major is computer science, and it was very hard for me to grab the information theory.
I found your video lectures on the information theory is the most understandable material for me.
I hope you could take some time to record the lectures on Rate-Distortion Theory.
I found your lecture note about it, but could not understand it without your explanation.
Send email to him.
@Cube Earth
I know this is old and you probably figured it out but he was basically the copying the first precision number of bits from the encoded message into z.
If precision is say 32, then whole is 2^32, when finding an approximation of z, if in the first iteration beta(i)=1, then Z = 2^32.... which means for any other 1 in the sequence Z will be more than whole. Am I misunderstanding? And if not, how can Z be in the range of [a,b)?
Hello, once I watched your intro to logistic regression (which was so great) but I cannot find it anymore, would you provide me with a link? Thanks so much!
*The pseudocode presented in this video contains one mistake that makes any implementation from it useless*
Here is my implementation of this algorithm in C++.
github.com/KyrietS/arithmetic-coding/blob/f80e7a911e898a45d114a1103149ad3b5feb0f40/AC_Core/src/AdaptiveScalingCoder.cpp
The structure is the same as presented in this video.
Look at the line '123' in my source code. THERE MUST BE A "BREAK" KEYWORD. I spent hours of debugging because of it...
I've also improved this example and made it more efficient using Binary Search for decoding symbols.
github.com/KyrietS/arithmetic-coding/blob/1584592fa666c55b3fd82b5af99afa5d9089743c/AC_Core/src/AdaptiveScalingCoder.cpp
You can find my newest version of Adaptive Scaling Arithmetic Coding on my master branch:
github.com/KyrietS/arithmetic-coding
Thank you
where are u ?
What is the M means? the length of the sequence bits? Thank you
yes
yes