₦airaland Forum

Sorry I have been offline guys. I have been writing my PhD thesis(still writing).

islamics:
Nice one! I was thinking you will classify your story into the three levels.
The witness will be the input layers.
The director getting the statement from the different witness will constitute the middle layer.
The CEO decision is the output layout.

P.S: I no too sabi but I guess the middle layer is where deep learning take place and where all those Mathematics are really involve.
Somebody should correct me if my assumption is wrong abeg.

True. I could have done that but I needed to emphasize a point. The point is that inputs are always features. Features can be learned (as in convolutional neural networks) or simply discretely extracted (as in artificial neural networks). Back to the example, you will agree with me that each of the six witnesses would have a perspective on what happened. Some may not have witnessed the entire proceedings. I can thus classify their perspectives as features (incomplete and partial observations of the actual event)

The best way to understand ANN is intuitively. The mathematics is beautiful only if you are mathematically oriented. Lets talk about a simple multilayer perceptron. Three levels; output layer, input layer and the middle layer. Let me use a generic example.
Two people fought in an office. 6 people witnessed it. As the CEO of the company, you set up a committee of directors to examine the facts. There are six directors in this committee. The goal is for them to examine the facts of the incident, they report the facts to you and you decide whether to fire the employees, suspend them or do nothing. So, to make this simple, the directors would talk only to the witnesses. Each witness, comes into the meeting room and narrates what happened (their perception of what happened), each individual director, tries to understand the events from the perspective of 6 different witnesses, he or she writes a report. You, as the boss reads each of the six reports and you come to decision; fire, suspend or nothing. This is a multilayer layer perceptron (ANN).
In reality, this example correlates to a pre-trained system, but this helps understand some of the underlying logic. I suggest you understand Logistic regression first, then the concept of classification. Then you move to a multi layer perception (forward training), then try to understand backpropagation(this is usually the hardest to get since is based on understanding how and why differential equations work. I hope this helps. Sorry I could not be of more help. I drew up a simple sketch to illustrate my story

Hello, I am interested in discussing and contributing. I'm currently doing a PhD in vision systems and machine learning in France plus I teach it to undergrads. I hope to have one or two things to contribute

I like this question. There are certain pitfalls however. How many grains are on the board?
1, 2, 3, 4, ..........
on the 64th square you would have 2 power 63 as many people have pointed out. But the real question is how many is on the board?
So you have to sum from the 0 to 63 using the geometric sum
(2 power (63 +1)) - 1

The minus one is not on the power, else you would only computing the number on the 64th square

In this article, the anisotropic diffusion filter used application such as Instagram and Snapchat is implemented on a step by step basis in python. A mathematical overview was first given, then a full fledged python implementation was given in two parts of the article
http://olumideokubadejo..fr/2016/07/anisotropic-diffusion-2-implementation.html

Sheggy13:
interesting research area. Perhaps we could talk more outside this place. I'm also in the academia. My email address is sheggswonder@gmail.com.. expecting yours. Cheers.

olumideokubadejo@gmail.com

Sheggy13:
Where are you doing your PhD and what are you working on bro?

France. Joint program between Universite de Grenoble and GeogiaTech. I am working on 3D segmentation and tracking

innarticles:
Hi lum1, I came over nairaland to find folks doing stuff in DS. I am DS enthusiast too. Done projects using LR, SVM, KNNs and also exploring CRNNs. Also very active participant at Kaggle.
I have some ideas to share. DM me if you find it convenient.

That is sweet. I am also on Kaggle but not too active. I have focused more in the past months on Image Processing and also currently doing a PhD which has taken almost all my time. I hope to get back to "Kaggle ways" soon lol. I would very much love to talk to you

foldl:
This sounds very much like Machine Learning to me.

Machine learning is knowing how to use implement statistical and probabilistic algorithms for use in Data analysis. Data science involves knowing what features use and train in this algorithms. There is not so much of a difference these days between both fields

ejihand1:
My brother, I have been following this thread since 3 days ago I saw it. Honestly, I don't seem to get why the thread was created. Is it a tutorial on data science or what. Op please you really need to come out and tell us your bearing as respect to this topic. Thanks.

Asalimbo, please take it easy on op. You might scare him off. I really want to learn from this thread.

I know it is pretty much too late to reply this post but forgive me. I have hardly been here. Well, I simply saw the thread and was very much interested that Nigerian Programmers were willing to talk about this topic since it sometimes borders on being too mathematical. We could translate it into a tutorial thread though. So data science is simply extracting information from data. Data can come in many formats; Text: A study was carried out from semantic analysis of Facebook posts and it was realised that the general mood of the population around december was sadness.Lol. Now, to automate the analysis, sentences have to be converted somehow to numbers. And analysed using some algorithm.

Data can also be an Image (which is where I specialise). Recognising faces, BIometrics etc. It is useful to know what features to collect and what algorithm to implement. The human eyes process 72 gigabytes of data a second. Our computers are not that fast. So we need to be able to extract only the relevant information necessary for the algorithms. It can also be speech. I hope I have been succinct enough. I would be happy to clarify more.

Wow..so sorry. I guess you should have applied from Abuja maybe

wildguy:
Okay that is good. I am supposed to be in Grenoble but my visa application is still being processed after 3 months. Thinking of deferring till next year.

I am doing a PhD. I suppose that is three years

wildguy:
Oh okay. That is good. Are you going for a one year course or two?

wildguy:
Hey lum1 have you been able to get your visa?

Yes I have. I am in Grenoble

What City are you in?

nosab:
Hello. For those of us now in France, whats been the experience so far? Any areas we are having issues maybe someone here can provide the solution or how they went about it?

Please seun, can you give the training set

Is this competition still on or a winner has been declared already

Obiwannn:
ERMMM.. OKAY BOSS. WHEN ARE U RESUMING AND WHEN DO U INTEND FIXIN UR VISA INTERVIEW?

I intend to do it this week since the minimum time to apply is 8 weeks before the start of the program. I would be starting October 1st

Obiwannn:
okay, got a scholarship to UJF...
The accommodation certificate is one which you get when you apply for a room. It shows thzt you have a room or are about to get a room and It should be signed and should contain your full duration of study period. I got mine directly from my school's accommodation office. Try doing one or two google searches to see if you can contact private landlords. They give this certificate too.

Thanks so much. You have been most helpful. We should hook up when we get there

No, Its INP

Hello,
I got a scholarship to Grenoble for a PhD . Fully sponsored. But, I am confused about the certificate of accommodation

Well, trust me. The best way to go about it is by using machine learning. Let the computer learn the difference on its own. Support vector machines, artificial neural networks and deep neural networks should be sufficient

Hopefully, my definition is close to what the OP has in mind but Data science concerns making sense of loads of data and effectively gaining insights into structures that these data relate to using computer science and statistics. For example, from data of how tumors have evolved to cancer in some people and how it did not in others, data science can define a general and fairly accurate predicting structure that would predict if a new patient with a tumor is likely to go on to have cancer. A study was done using data science on Facebook and algorithm was able to tell the emotional state of facebook users. It showed that facebook users were on an emotional low during the end of the year because most break-ups occur around that time. With data science and Machine Learning, it is possible to create systems that learn from data. Support Vector machine and Neural Networks are techniques by which learning can occur in data.

Data science is a pretty interesting field. Could be complicated when you are at the algorithmic level. I only do research in this field. I utilize learning algorithms to extract information from datasets and generate models. I mostly use neural Networks and Support Vector machines but I am gradually switching to deep neural network models.

all4naija: Lol... I can see you have movement simulation project done in Python. But, it is only apply to the linear form of movement and I quite agree with what you are doing is a linear but there is no idea for different movements to choose from among many thereby can be applied to other fields like robotic movement and many others. You have to see beyond mere algorithm of calculating a distance.

Lol...I seriously do not understand what you just said. I am not simulating motion, I am recognizing motion and in essence using it as a biometric. Using silhouette based approaches and model based approaches. I cannot upload the video ( it would be a breach) so I uploaded a single frame and I am sure you know that we have about 25 frames per second. I also do not understand how human motion is linear. If you take a model of hip rotation and thigh angle rotation during motion it would reveal a "Pendulum-like" motion and with the little I know it does not seem linear..Forgive me if I am wrong. The algorithm being used is.. I am callculating the change in symmetry as motion is in progress. The unique change in symmetry is getting a 93 percent recognition rate. The second algorithm without any markers calculates hip and thigh rotation angles. That also is unique.

elvis10ten: I was about asking you what you used to implement the algos. can we see a picture or video of your work in action ? or a paper on it ? or anything to show what you just said, you know talk is cheap.

I should not really be doing this but attached is a screenshot of the silhouette of a walking subject during the processing and recognition phase. Would be writing the research paper in the coming weeks

Attended FUT Minna (even though very little was learnt here). Went through the route of "Learn Yourself". I hate to keep tabs on programming languages cos I see it as a means to an end and not an end in itself. I love artificial intelligence and Robotics as there is no limit to what you can do. Currently studying it at Southampton and my research is majorly in Biometrics; Just developed a new way of recognizing humans by the way they walk and hoping to do some research into building computer algorithms that recognize emotions and interact with you based on your emotional state. Even though I hate keeping tabs on Programming Languages, I love C (for Embedded systems) and Python. If you really want to learn computer science in Nigeria, get the books and read them.

workinprogress: i also encountered this problem...what i did was allow the appointment date pass then i filled another online form and used the new GWF number to book my new appointment and i got my visa successfully....so thats the only solution i have for u....i emailed vfs for days and they kept giving me stories so i did what i wrote above....ALL THE BEST.....after making the new appointment make sure u record the reference number on that page or download the pdf cuz once u leave that page u cant get the reference number again

thanks so much