Zinnia Muntaha Mowri

General: This is the home page of a channel. It is similar to someone's profile "wall" in a social network context. Posts created by the channel are displayed according to the observer's viewing permissions.
Create a Post: If you have permission to create posts on the channel page, then you will see the post editor at the top.

QAM modulation+AWGN+NB-LDPC decoding

Wed, 31 Jul 2024 17:08:20 -0600 last edited: Wed, 31 Jul 2024 17:08:57 -0600

Hello Dr. @Winstead,

After some parameter adjusting in the algorithm that I am working with (16QAM modulation+AWGN+NB-LDPC decoding), I got the attached figure where the coded SER is slightly better when SNR was 12.5 to 16 db. So, it seems like the decoder started working, and this is the figure for 20 iterations only. I will run the simulation for more iteration and see if it shows any difference.

Also, I am trying to improve the strategy to find the best candidate to replace the erroneous symbol (not complete yet).

For now, the decoder is working based on finding symbol with nearest Euclidean distance.

I am attaching the figure, if you have any insight seeing the curves, let me know.

Thnak you.

Zinnia Muntaha Mowri

Mon, 19 Aug 2024 15:56:39 -0600 last edited: Mon, 19 Aug 2024 16:00:40 -0600

Hello Dr. @Winstead, I tried to change the parameters value theta and eta and got some good performance of the decoder. It seems like changing the parameters after certain SNR improved the performance and helped to escape from the error floor region. I kept the parameters as below:
For SNR < 11, theta = -2.5; targetFER = 10; eta=1.1;
For SNR >= 11.5, theta = -1.9; targetFER = 2; eta=1.3;

Zinnia Muntaha Mowri

Fri, 23 Aug 2024 15:49:36 -0600 last edited: Fri, 23 Aug 2024 17:37:32 -0600

Hello Dr. @Winstead

I have compared the decoding performance (in terms of Bit Error Rate, BER) of the decoder I am developing with that of the min-sum decoder. I found a reference paper that presents the simulation results for the min-sum decoder's BER performance using GF(16)-LDPC codes under an AWGN channel with 16-QAM modulation; coding rate of 1/2 and a code length of 2016 bits (equivalent to 504 GF(16) symbols). I extracted approximate data points from the paper's result figure 1 and plotted them (the blue line). For my noisy symbol flipping decoder (orange), the parameters are similar, with a coding rate of 1/2. However, the code length in my implementation is 816 bits ( 204 GF(16) symbols). And I have decoded zero codeword, for other codewords, there may be some problem in my coding, or may be I could not figure out best simulation parameters yet.

The paper link is here and the simulation result is attached.
#^https://ieeexplore.ieee.org/abstract/document/4595129
Min-Max decoding for non binary LDPC codes.
2008 IEEE International Symposium on Information Theory.

CRN math problem

Zinnia Muntaha Mowri

Wed, 29 May 2024 11:35:11 -0600

Dr. @Winstead attached you will find the document where I have articulated the notes on the CRN math problem.
Let me know what else I can include to improve the idea and also if I am missing any important thing.

I have not shared the document with Josh and Dr. Zhen yet. Once you give me feedback on that, I can revise and share with them.

FLUENT research

Winstead

Wed, 29 May 2024 12:05:09 -0600

I think it looks good. A technical point: the condition P02*P20 < 1 is guaranteed, since we are using probability quantities they are always between 0 and 1. An aesthetic point: how did you create the figures? They look a little pixelated, it would be better to use a vector image format.

Let’s meet in the lab today (not my office). I’ll meet you there around 1:15.

Zinnia Muntaha Mowri

Wed, 29 May 2024 12:18:34 -0600 last edited: Wed, 29 May 2024 12:34:52 -0600

I created the figures in drawio.com online and exported them as .png file and used them in the overleaf project. I also noticed that the figures are not very clear. I will replace the figures, but I have to learn the technique to draw in vector image format. I am trying to learn and draw those figures in TikZ, because I remember you mentioned about TikZ.

Zinnia Muntaha Mowri

Thu, 30 May 2024 11:48:50 -0600 last edited: Fri, 31 May 2024 10:27:48 -0600

@Winstead I have redrawn the diagrams with TikZ pictures. Attached is the updated document.

A comprehensive study of mobile-health based assistive technology for the healthcare of dementia and Alzheimer's disease (AD)

Zinnia Muntaha Mowri

41.751321 -111.8076158, Wed, 03 Apr 2024 05:44:43 -0600 last edited: Wed, 03 Apr 2024 05:52:33 -0600

#^https://link.springer.com/article/10.1007/s10729-019-09486-0

This paper highlights mobile health (mHealth) apps developed to assist people with dementia (PwD) and their caregivers with different purposes such as tracking, dementia screening, caregiver guidelines, reminiscence, and socialization.

-- Regarding the involvement of medical teams in the validation or testing of these apps, the paper suggests that healthcare professionals and physicians should be involved in the development of apps. But does not confirm the involvement of professionals for each app, especially for the apps that determine the possibility or predict dementia in old age. In Table 6, they have listed some of the app's names, advantages and disadvantages, and privacy policies but I did not find any explicit information about the involvement of the medical team in the design process. In that way the credibility or reliability of the apps for screening is questionable. They have argued that mobile-based dementia screening can be used as primary screening at home, replacing paper-based clinical tests due to their capability of displaying immediate results and maintaining uniform scores and history. However, they highly recommended that these kinds of app-designing processes should involve medical professionals. It is suggested that "smartphone technology has the capability to make behavioral health care more efficient and interactive for patients, but healthcare professionals should also assess the risks involved in using this technology and they should use the clinical judgment to consider when and for which of their patients need technology-based service [80, 81]."
Some of the apps seem to be validated or tested by the medical team, such as the decision support app [74] helps professionals in making decisions and judgments, and seems that they have involved the medical team in the development process, though no clear evidence was provided.

-- While this paper broadly categorizes mobile apps, few wearable devices for assisting the caregiver have been mentioned. Such as Caho et al [72] used smartwatches and smartphones for caregivers. The smartwatch fetched the GPS location of the patients and sent it back to the smartphone via Bluetooth. Probably this was used to find nearby public healthcare centers. This paper also mentions that further research should be conducted in tracking of dementia patients -- it seems that there is still a gap in tracking or monitoring AD patients or tracking the total number of patients for recording the increasing percentage. It seems that this study only targeted smartwatches, and mobile or tablet-based assistive technology developed from 2014 to 2019. They have mentioned that further research is needed to investigate the usage of other smart devices such as smart TV or smart home or other assistive technologies (computer or sensor-based).

ADRC

Zinnia Muntaha Mowri

41.75134968759645 -111.8076562380199, Mon, 25 Mar 2024 08:51:09 -0600 last edited: Mon, 25 Mar 2024 08:56:42 -0600

Dr. @Winstead, I won’t be able to come today for our meeting. Can I come on Wednesday 1 pm during your office hours?

Eric Reiss

Mon, 25 Mar 2024 09:55:19 -0600

@Winstead I also just logged in to say that I need to run my wife to her doctor appointment, so I also won't make it in today. I could do Wednesday from 1pm to 1:30pm.

Winstead

Mon, 25 Mar 2024 10:22:57 -0600

Okay, Wednesday should be fine

Regarding Comprehensive exam questions and some doubts

Zinnia Muntaha Mowri

41.7513267 -111.8076324, Wed, 21 Feb 2024 13:57:49 -0700 last edited: Wed, 21 Feb 2024 14:01:46 -0700

Dr. @Winstead I tried to put my doubts and thoughts here.

Paper: B. Amiri, J. Kliewer, L. Dolocek, Analysis and enumeration of absorbing sets for nonbinary
graph based codes, IEEE Trans. Commun., vol. 62, no. 2, 2014.

Question 1.
answer attached from the solution you have provided.

Doubt: I think the answer - Every trapping set is an absorbing set -- should be the other way around according to the paper. Because for trapping sets-- the number of satisfying checks doesn't need to be greater than the number of unsatisfied checks connected to each variable node.
Whereas, to be an absorbing set, the number of satisfying checks needs to be greater than the number of unsatisfied checks connected to each variable node. |E|>|O|
These conditions I interpret as absorbing sets are special types of trapping sets. I may conclude that every absorbing set is a trapping set, but not every trapping set is an absorbing set. Which contradicts your solution.

Please correct me if I am understanding this in the wrong way.

Zinnia Muntaha Mowri

Wed, 21 Feb 2024 14:22:35 -0700 last edited: Thu, 22 Feb 2024 12:01:06 -0700

I am trying to explain my second doubt here.

Question 3.
answer attached from the solution that you provided.

Doubt:

the number of rows of Matrix B is the number of satisfied checks and the number of rows of matrix D is the number of unsatisfied checks. Now each variable node from the set V = (v0, v2, v3) should be connected to more satisfied checks than unsatisfied checks.

The last line of the answer "The second condition concerns the relative “weight” of elements in D versus those in B. For every column, there are more non-zero elements in B than in D, so the condition is
satisfied." --- I think there is a little issue--- because (column 2 and 3 in matrix B) v2, v3 have one non zero element in matrix B. v2 and v3 have one non-zero element in matrix D also. So, for v2 and v3 number of connected satisfied check nodes is not greater than the number of connected unsatisfied check nodes. Actually for v2 and v3, |E| =|O|.
In that case, I assume that for every variable node, the second condition is not met, and therefore the set of variable nodes V = (v0, v2, v3) is not an absorbing set.

Also in this paper they mentioned in remark 1, that they are considering each variable node to be
connected to more satisfied checks than unsatisfied checks to be an absorbing set.
This differs with other paper they referred to --- where they consider more satisfied checks than unsatisfied checks, taken collectively over all variable nodes. So that is the difference they are talking about.

Remark 1. "Our definition of a non-binary absorbing set
is different from the definition proposed in [29] where the
authors define a non-binary (primitive) absorbing set as an
object which has more satisfied checks than unsatisfied checks,
taken collectively over all variable nodes in this object. In
contrast, our definition, similar to the original definition of
binary absorbing sets [13], require each variable node to be
connected to more satisfied checks than unsatisfied checks. "

so, I think the set of variable nodes V = (v0, v2, v3) is not an absorbing set. Please let me know if I understanding it wrong.

Winstead

Thu, 22 Feb 2024 14:10:32 -0700

I think these are great items to discuss in your oral examination. Be prepared to present these points to the committee, you may need to review the definitions and notation so the committee can understand your arguments.

Zinnia Muntaha Mowri

Thu, 22 Feb 2024 14:16:00 -0700

Okay thank you so much for your guidance.

Symbol flipping algorithm

Zinnia Muntaha Mowri

41.7523158 -111.8203296, Tue, 06 Feb 2024 15:44:07 -0700 last edited: Tue, 06 Feb 2024 15:44:58 -0700

Dr. @Winstead, I have tried implementing the symbol-flipping algorithm over GF(4) for a given parity check matrix in octave. One approach I tried was checking the syndrome for each symbol in the received code word. Another one is by considering hamming distance... min hamming distance between the received vector and valid codewords resulting in the corrected codeword.

I was not able to commit the files in our LDPCsimulaiton repo. Hence, I uploaded them in the below repo of mine. I have not added a README file yet, which I will do shortly. There are some issues in the code or algorithm that I might need to discuss with you tomorrow.

#^https://github.com/zinniamowri/symbol_flipping_GF4

show all 7 comments

Winstead

Wed, 07 Feb 2024 09:41:25 -0700

I'm looking at your code now but won't be able to meet today. We can discuss the issues via comments here if that helps.

Zinnia Muntaha Mowri

Wed, 07 Feb 2024 10:25:03 -0700

Yes, I am trying to explain my question here.

For the given parity check matrix the valid codeword list: 0 0 0, 1 2 1, 2 0 2, 3 2 3.
Both the code symbolflipping.m and hammingDistance.m can correct single-bit errors. Suppose, the received codeword is 2 2 3, they can give the corrected vector 3 2 3 (both symbolflipping.m and hammingDistance.m). But certainly, if there are multiple-bit errors symbolflippig.m can not find any valid codeword.

Whereas, hammingdistance.m gives a valid code word in return, but can be another valid codeword instead of the current one. Suppose, the received codeword is 1 3 2, the valid code word is returned as 1 2 1. This will be correct if the 2nd and 3rd position is erroneous. But what if the 1st and 2nd position was erroneous, in that case, the correct should be 2 0 2, and the returned value is wrong.

So multiple-bit error correction is not working. Is there any way to analyze the syndrome for all possible received vectors and locate the erroneous position exactly? For that, I have written a script com_all_rec_vec.m to get the corresponding syndrome for all possible received vectors, if that can be helpful. But the syndromes are not unique, so I don't know how to utilize them.

Winstead

Wed, 07 Feb 2024 12:26:16 -0700

Usually a non-iterative syndrome decoder can only correct a single symbol error, and only if the code's structure is such that the non-zero syndrome corresponds one-to-one with the single error. When multiple errors occur, the syndrome method may fail to detect them properly and may return a false codeword. The latter case is called an undetected error, since the decoder's output is presumed correct when valid.

In the iterative symbol flipping algorithm(s), the syndrome is computed in each iteration, and is used as one part of the per-symbol reliability metric. In single-symbol-flipping, the position with the "worst" reliability metric is "flipped" (altered) so as to improve its reliability. After the flip, the syndrome must be recomputed along with all reliability metrics.

If I understand correctly, your code is implementing a non-iterative hard-decision syndrome decoder. Is that right?

Zinnia Muntaha Mowri

Wed, 07 Feb 2024 12:33:41 -0700 last edited: Wed, 07 Feb 2024 12:36:47 -0700

Yes it is non-iterative. Thank you for your comment. I will take the symbol-flipping paper and try to implement the original symbol-flipping algorithm (iterative).

Zinnia Muntaha Mowri

Wed, 14 Feb 2024 12:04:09 -0700 last edited: Wed, 14 Feb 2024 12:06:52 -0700

I have committed two-octave codes in the repo #^https://github.com/zinniamowri/symbol_flipping_GF4. Trying to implement an iterative symbol flipping following this paper ( Weighted Symbol-Flipping Decoding for Nonbinary
LDPC Codes, #^https://ieeexplore.ieee.org/abstract/document/5480573) cited in the randomly penalized symbol flipping paper. Because it seemed to be comparatively easier to start with.

The script named calculate_E_n_k.m in the repo, calculates the flipping metric value of each symbol in the received codeword and returns the index of the symbol that should be changed.
But to which symbol it should be changed into, yet not completed in the code. I am working on it.

Separately, the script named mappingToBits.m works for mapping the GF field elements to their corresponding bit representation, BPSK modulation, adding noise, and returning a hard decision symbol vector as y. So y is again the received codeword representation in the GF field. These functions I would need when the whole algorithm structure is ready.

Every variable in the script I tried to follow as the notations in the paper.

Zinnia Muntaha Mowri

Fri, 29 Mar 2024 10:10:04 -0600

Dr. @Winstead, I have committed the code that I was showing in the meeting. The calculate_E_n_k.m file is where I calculate the flipping metric using the LLR and other terms according to equation 8 in the paper (the paper is also in the repo). I have also committed the LLR.m file where I calculate the LLR for each symbol, I just committed it separately to look into it more easily. But essentially the LLR calculation in calculate_E_n_k.m is the same as in LLR.m

#^https://github.com/zinniamowri/symbol_flipping_GF4

Zinnia Muntaha Mowri

Wed, 03 Apr 2024 05:55:55 -0600 last edited: Wed, 03 Apr 2024 05:56:40 -0600

I have added a header in calculate_E_n_k.m script including the paper reference and some description of the notation or variables used. Also, a main.m script is separated out from the previous code. So now the calculate_E_n_k.m script only contains the function definition and main.m script calls that function.

Server Rack Measurement

Zinnia Muntaha Mowri

41.751322 -111.8076169, Fri, 10 Nov 2023 11:07:38 -0700 last edited: Fri, 10 Nov 2023 11:13:08 -0700

@Winstead and @Eric Reiss Yesterday I measured the dimensions of the server rack. I think if we buy a Rack Shelf of 19-inch length and 14-inch depth, that would work. The holes are of around 5 mm in diameter, but I need to measure them again with a smaller scale, which I did not have yesterday. But I think a 10/32 rack screw may fit because I found that 10/32 rack screw diameter is 3/16 inch or 4.76 mm. In the comment, I have shared an example rack self that I found on amazon.

Zinnia Muntaha Mowri

Fri, 10 Nov 2023 11:08:46 -0700 last edited: Fri, 10 Nov 2023 11:15:56 -0700

#^https://a.co/d/bey8xfr

AC Infinity Vented Cantilever 1U Universal Rack Shelf, 14" Deep, for 19" equipment racks. Heavy-Duty 2.4mm Cold Rolled Steel, 60lbs Capacity

#^https://a.co/d/9ioh1qu

Eric Reiss

Fri, 10 Nov 2023 13:45:23 -0700

Ok, I talked to Rory, and he said we could just send him a link with the index to use. @Winstead I remember you saying something about indices at our last meeting. Do we have an index to use for that? Rory said that if we don't have one, we might be able to have Wiley order it under IT.

Winstead

Tue, 14 Nov 2023 08:23:19 -0700

We found the account indices, I will forward the correct number to Rory. Thanks for doing the measurements.

proposal update

Zinnia Muntaha Mowri

41.7513375 -111.8076309, Mon, 25 Sep 2023 17:08:01 -0600

@Winstead Attached is the pdf of the proposal where I am trying to update some recent works or literature reviews on the non-binary LDPC code decoding. Please see the Background section, prior work subsection, on page 5, second last paragraph starting with "It has been well demonstrated that very long binary LDPC codes........."
Please let me know how we can make a better flow in this section to build the background.

show all 10 comments

Winstead

Mon, 02 Oct 2023 13:09:51 -0600

Don't forget to send me your modified source file so I can setup a repository for it.

Zinnia Muntaha Mowri

Mon, 02 Oct 2023 15:58:30 -0600 last edited: Mon, 02 Oct 2023 15:59:35 -0600

Oh thank you, I forgot to do that earlier, sorry. Here I have attached the source as a zipped folder. Let me know if it works.

Zinnia Muntaha Mowri

Mon, 09 Oct 2023 21:31:50 -0600

I have updated and committed some changes in the proposal with the recent Syndrom Bit Flipping Decoding paper content. Also deleted some of Vasic's related sentences and all other collaboration-related areas, as much as I understood clearly.

Zinnia Muntaha Mowri

Fri, 13 Oct 2023 14:45:39 -0600 last edited: Fri, 13 Oct 2023 15:30:05 -0600

@Winstead I was wondering if I need to install package pandoc-xnos separately? Because I tried to do a make and got this below error. Another thing is, when I will do a make, will it generate the pdf or I have to do it from emacs (tool->compile)? Please let me know what should I do to solve this issue.

Error:
[zinnia@genesys nsf_ldpc]$ make
pandoc --filter pandoc-xnos --citeproc --verbose --template=template.latex proposal.yaml proposal.md -s -o proposal.tex
[INFO] Loaded template.latex from template.latex
[INFO] Running filter pandoc-xnos
Error running filter pandoc-xnos:
Could not find executable pandoc-xnos
make: *** [proposal.tex] Error 83

And when I compiled in emacs, I get the below error:

make -k
pandoc --filter pandoc-xnos --citeproc --verbose --template=template.latex proposal.yaml proposal.md -s -o proposal.tex
[INFO] Loaded template.latex from template.latex
[INFO] Running filter pandoc-xnos
Error running filter pandoc-xnos:
Could not find executable pandoc-xnos
make: *** [proposal.tex] Error 83
make: Target `proposal.pdf' not remade because of errors.

Compilation exited abnormally with code 2 at Fri Oct 13 17:24:59

Zinnia Muntaha Mowri

Mon, 16 Oct 2023 22:19:38 -0600

Motivation example 3 has been added based on the result found in the recent SBF paper. Figures yet to be added.

Winstead

Tue, 17 Oct 2023 11:46:02 -0600

@Zinnia Muntaha Mowri pandoc-xnos is a separaten python package that can be installed using pip or pip3. Are you building on genesys?

Zinnia Muntaha Mowri

Tue, 17 Oct 2023 11:47:31 -0600

@Winstead yes I tried to build in genesys.

Winstead

Tue, 17 Oct 2023 13:27:18 -0600

@Zinnia Muntaha Mowri pandoc-xnos should now be available to all users.

Zinnia Muntaha Mowri

Tue, 17 Oct 2023 13:38:02 -0600 last edited: Tue, 17 Oct 2023 13:48:33 -0600

Okay thank you, I will try to build and let you know.

Zinnia Muntaha Mowri

Tue, 17 Oct 2023 13:49:17 -0600

The build was successful and pdf file is generated. Thank you so much.

Unable to attend today's meeting

Zinnia Muntaha Mowri

41.7513277 -111.8076239, Tue, 12 Sep 2023 10:25:49 -0600

@Winstead My son had a fever for a couple of days and last night he had a high fever. I am sorry that I am not coming to our weekly meeting today. I will try to catch up about the meeting discussion with Eric or you later this week.
Thank you.

show all 4 comments

Zinnia Muntaha Mowri

Mon, 18 Sep 2023 20:17:34 -0600 last edited: Mon, 18 Sep 2023 20:18:06 -0600

@Winstead I very sad to inform you that I am affected by the virus fever very badly. My son has not recovered yet and now me and my husband got attacked by the virus. Right now I have 102.9 degree Fahrenheit fever. In this situation I don’t think I am able to join tomorrow morning. Again take my humble apologies.

Winstead

Tue, 19 Sep 2023 09:12:54 -0600

I hope you recover, that is quite a high fever. You may wish to consult a doctor.

Zinnia Muntaha Mowri

Wed, 20 Sep 2023 17:42:46 -0600

@Winstead thanks for your advice. I went to doctor today morning, she is suspecting it might be covid, or influenza or pneumonia. I did covid test at home yesterday and it was negative. But doctor said it might be false negative, no guarantee. She suggested me to do a test, chest X-ray. I will go for the test tomorrow. Thanks.

Stratified sampling related question

Zinnia Muntaha Mowri

Wed, 30 Aug 2023 13:05:16 -0600 last edited: Wed, 30 Aug 2023 13:30:36 -0600

My question was, if the probability of choosing a grid interval is uniform, and the width of the interval is uniform, then how the thinner width intervals will produce higher probability of samples? I found the answer by watching the part 2 tutorial again.
I found that this is based on based on VEGAS algorithm and adapting IS.
In VEGAS algorithm, first we are dividing the domain with N intervals of size x (width of the interval). Probability of choosing one interval is 1/N. Within the interval, choosing a uniformly distributed sample point.
So, Probability of choosing a sample from a particular interval is:
P = (1/N)(1/x).
and it is inversely proportional to to the width of the interval x. So with thinner or lesser interval/bin will have higher probability, as the number of intervals is same N. When adapting technique is applied, the grid interval is adjusted (changed) with time towards the optimal probability distribution which is g*(H)=p(h)/w(h),
where w(h) is the weight function. So by adjusting the interval width, narrow interval will produce higher incidence of samples.

Sorry, if I am not able to make it clear here. I can explain it when wee meet in the next meeting.

Winstead

Thu, 31 Aug 2023 12:41:52 -0600

I believe I follow. If you produce an equal amount of samples per interval on average, then the narrowest intervals will have the highest density of samples. The probability density is approximately S/dx, where S is the sample count per interval and dx is the interval width.

Git branch creating issue

Zinnia Muntaha Mowri

Thu, 24 Aug 2023 12:19:30 -0600 last edited: Thu, 24 Aug 2023 12:29:26 -0600

@Winstead I think I am unable to git push anything in the fluentverificaiton git repos. I tried to create a branch but seems like I do not have the permission.
I also tried to commit a empty text file in test basis, but I could not git push it because of bellow error. Is there any write access that I need to have before I can commit any changes in the repo? Although, I have configured my email and user name.

fatal: Authentication failed for 'https://github.com/fluentverification/usu_importance_sampling.git/'

Zinnia Muntaha Mowri

Thu, 24 Aug 2023 15:14:13 -0600

@Winstead thank you, it is working now. I am able to commit changes.

FLUENT project

Zinnia Muntaha Mowri

Tue, 22 Aug 2023 11:10:09 -0600

@Winstead I have went through the readme file and tried to run some code by myself to understand the code flow.
./gradlew build (what is this command actually doing I do not know but it worked)
I suppose the prism is already installed in our genesys server. Nevertheless, I tried to run this command ./bin/initPrism.sh
to install prism. It was rejected and I got some msg to be more careful to what I type! I realized that it was the part of the troubleshooting.

I gave the command like: Where I think M is the multiplier to each state in the catalog if the path make the way to the target in the given iteration.
./bin/run.sh simulate -M 2
output: Probability to reach final state: 0.02201247311827907, Variance 5.726917899717129E-7, non-zero samples 839, useful sample rate 0.00839 M=2.000000 TMAX=1000.000000 Nruns=100000 modelFile=models/example.pm

If can get more insight about the output, I can add some details in the readme file with example. I think this can help the user.

second I run the bellow command for model generation I guess.

./bin/run.sh mgen sparse --output sparse-model2.prism

I can see the file sparse-model2.prism is created with 10 states. But I do not know what the seed is meaning and also other syntax are not clear to me. If I can get some brief on the files, that would be helpful.

Also, /Senior_Project/Tom_SP/ this folder containing some files seems nothing it it. Should I clear that up?

Zinnia Muntaha Mowri

41.7669718 -111.8551201, Fri, 11 Aug 2023 14:22:15 -0600 last edited: Fri, 11 Aug 2023 14:24:02 -0600

@Winstead I forgot to ask, if you are available on Wednesday, Thursday or Friday next week. Next week Tuesday, 15th August is my last day on CPT, so I wanted to work on Monday and Tuesday in my internship. So if we can meet on Wednesday or any later day next week that would be good, let me know a suitable time of yours.

Winstead

Mon, 14 Aug 2023 13:26:46 -0600

Thursday is most likely; I have Department/College/Research retreat meetings all week, but the College retreat only runs through 2pm.

Zinnia Muntaha Mowri

Mon, 14 Aug 2023 18:27:25 -0600

Okay I can come on Thursday after 2.

Work hour plan for Fall 2023

Zinnia Muntaha Mowri

41.7670092 -111.8550807, Thu, 10 Aug 2023 15:38:45 -0600 last edited: Thu, 10 Aug 2023 15:40:23 -0600

Here is my proposed work hour plan per week.
Research projects = 10h (Alzheimer project and the other one where Eric is working, 5h each)
TA hour = 10h

and I don't have specific salary expectations, but I used to get $18 per hour, so if I can earn that much, it would be good.

@Winstead If you have any suggestions for the work-hour plan, I can work accordingly.

Zinnia Muntaha Mowri

Fri, 11 Aug 2023 15:37:32 -0600 last edited: Fri, 11 Aug 2023 15:38:19 -0600

@Winstead I just got an email from Tricia that I will not owe anytime to the department, hence my tuition will be paid from your research projects, and in that case, I can work the entire 20h on the research projects.

Winstead

Mon, 14 Aug 2023 13:25:33 -0600

I think we can split your time equally between the two projects, on average.

Zinnia Muntaha Mowri

Mon, 14 Aug 2023 18:26:31 -0600

Okay thank you.

Non_binary_LDPC_decoder

Zinnia Muntaha Mowri

41.7759232 -111.8011392, Tue, 01 Aug 2023 03:15:37 -0600 last edited: Tue, 01 Aug 2023 03:20:34 -0600

Attached you will find the slides on non_binary_LDPC code. I have tried to make slides on what I have figured out by reading the non-binary LDPC decoding papers so far. I am still working on the slides to make the correct subscript notation, please ignore that for now. Also, I will work on the part about how I can add those references as we do in LaTex. I have used MARP (for Windows, extension available in VS code) to make these slides.
I found the templates for markdown slides here: #^https://left.engr.usu.edu/slides/
But unfortunately could not find those in genesys, so could not make it through the templates. We can try to figure out that when we will meet.
For now, please have the attached HTML file and give your comments.

Slides

Apologies for begin unable to meet today

Zinnia Muntaha Mowri

41.7513257 -111.8076125, Tue, 25 Jul 2023 10:16:51 -0600

Dear Dr. @Winstead, I am sorry that I won't be able to come to today's meeting due to some family emergency. I will try to catch up next week, sorry about that.

Winstead

Tue, 25 Jul 2023 14:25:46 -0600

Okay.

test on comments

Zinnia Muntaha Mowri

41.7670218 -111.8551507, Thu, 20 Jul 2023 15:31:18 -0600 last edited: Thu, 20 Jul 2023 16:50:45 -0600

@Winstead I have cross-checked the comment disable option, it is off now. Let me know if you can comment here in this post. Also I think you might forget to upload the notes from our last meeting. It would be very helpful if you can upload the images.

Thank you!

testing

Winstead

Tue, 25 Jul 2023 14:24:52 -0600

Yes I can comment here.

Recursive sum operation

Zinnia Muntaha Mowri

Tue, 18 Jul 2023 11:55:53 -0600 last edited: Wed, 19 Jul 2023 14:04:21 -0600

Last meeting we were trying to find out why the recursive sum operation is not a Galois field.?
I think that the recursive sum operation we tried to prove on integers and integers does not form a field because most of them do not have multiplicative inverse except {1,-1}. And this is an infinite field. To be a Galois field, it needs to be finite. I tried to prove the recursive sum on rational numbers (multiplicative inverse exists) and all the rules to be a field work but rational numbers are also infinite fields.
And for addition and multiplication operations we have the inverses, 23 has additive inverse -23. 23 has multiplicative inverse 1/23.
But what about the recursive inverse, 23 will it be -1 or +1 or -5! That I don't know.
Another thing is that the finite field of GF(q) does not exist for all q. q must be equal to p^m, where p is prime and m is a positive integer.

So these are my assumption why this is not a Galois field.

Galois field

Papers

Zinnia Muntaha Mowri

Thu, 13 Jul 2023 00:27:04 -0600 last edited: Thu, 13 Jul 2023 01:00:31 -0600

The below 3 papers are on non-binary LDPC code and were published at the ISIT conference. We may look into the details for some new ideas to be included in the proposal.

1. Performance of ML Decoding for Ensembles of Binary and Nonbinary Regular LDPC Codes of Finite Lengths
#^https://ieeexplore.ieee.org/abstract/document/8006637
This paper estimates the ML decoding error probability for the AWGN channel by using average weight enumerators for both binary random LDPC codes and binary images of random nonbinary LDPC codes. The new upper bounds are obtained on the ML decoding performance, based on the precise coefficients of the average weight spectra. However, they have concluded that ML decoding performance should not be used as a target for searching for good iteratively decodable codes.
2. Construction of Rate (n - 1)/n Non-Binary LDPC Convolutional Codes via Difference Triangle Sets
#^https://ieeexplore.ieee.org/abstract/document/9174510
This paper analyzed the structure of different triangle sets to construct non-binary LDPC convolutional codes whose parity check matrices are free from 4-cycles and 6-cycles. They have given the construction of rate (n - 1)/n convolutional codes over non-binary fields, generalizing a construction from Robinson and Bernstein, using difference triangle sets.
3. A Class of Non-Binary Doubly-Generalized LDPC codes for Moderate and High Code Rates
#^https://ieeexplore.ieee.org/abstract/document/9518035
They proposed a small fraction of single parity-check codes at the side of the variable node. They have claimed that (together with the use of extended alphabets) the existence of such a fraction has been shown to improve the asymptotic decoding threshold, without harming the minimum distance.

Summary

Papers

Zinnia Muntaha Mowri

Tue, 20 Jun 2023 12:43:43 -0600 last edited: Tue, 20 Jun 2023 12:55:52 -0600

In this paper "A survey on channel coding techniques for 5G wireless networks" they have mentioned the Issues addressed by different researchers and future directions in Channel Coding field (section 4) and applications (section 6). Research challenges faced in the implementation of these techniques in terms of encoding or decoding, error floor, security, cycles impact on decoding, security and energy efficiency.
I can look in that section in details and some viable references for our proposal. May be I can look into such papers 2,3,4.

I don't know if the journals where they are published are good enough or not but from these paper we can get other references and future directions for our proposal.

1. A survey on channel coding techniques for 5G wireless networks
#^https://link.springer.com/article/10.1007/s11235-019-00630-3#Sec19

2. Set Message-Passing Decoding Algorithms for Regular Non-Binary LDPC Codes.
IEEE Transactions on Communications
#^https://ieeexplore.ieee.org/abstract/document/8017555

3. A Parity Structure for Scalable QC-LDPC Codes With All Nodes of Degree Three.
IEEE Communications Letters
#^https://ieeexplore.ieee.org/abstract/document/7938596
claiming that:
A novel block triple-diagonal (BTD) parity structure with all nodes of degree three is proposed for quasi-cyclic (QC) low-density parity-check (LDPC) codes. The BTD structure without low node degree can effectively improve the error floor performance and, meanwhile, enable low complexity encoding.

4. Fifth generation (5G)cellular and the network for tomorrow: cognitive and cooperative approach for energy savings
#^https://www.sciencedirect.com/science/article/pii/S1084804516303125

paper list