From OpenCV and Tesseract to exploring recent research results in Computer Vision [Challenge 2]

From OpenCV and Tesseract to exploring recent research results in Computer Vision [Challenge 2]

challenges, Computer Vision, Deep Learning, learning, Technology

I’ve been away for a while… Actually, I didn’t notice as the time was going on.. even though I felt like I’ve been running all days since my last article 😀 .. I was all in a new dimension of machine learning..one that is so huge and with many many possible doors to knock!

-What is it that you called “new dimension”?

Okay, let me tell you. If you followed my previous post about OpenCV and Tesseract, you may already know that I’ve been working on a computer vision project, involving Object Character Recognition (OCR) for texts in a natural scene image – By the way, I really like the problem that I’m trying to solve, that’s why I accepted working on the project despite the fact that I don’t have a previous knowledge of the topics and technologies used in such contexts. –

Back to the subject 🙂 .. so after trying out with tesseract and openCV, I found that there are many parameters that can affect the result, these parameters relate especially to a phase that it’s emphasized from experienced people.

What is that phase?

The preprocessing phase of the images. What I mean by preprocessing is the preparation of the input data (in this case, a natural scene image containing text) in a way that allows the algorithm to deliver good results.

How is that? preparing an image? What can someone do after getting the image?

Yes preparing it! It’s like when you want to do a cake, you have to prepare its ingredients.. or when you want to teach your child to differentiate between dogs and cats, you have to collect photos of cats and dogs beforehand, that will be your material in the process.

And you can prepare an image by playing around the color intensity of each pixel, regions, contours, edges, alignment.. all these are image properties. These properties are what make variations present in an image e.g blur, noise, contrast, etc. Adjusting these properties to control variations is necessary in order to identify(detect) text regions, segment these regions correctly and recognize the text as a last phase in the pipeline.

I think that at this moment you are starting to understand why it’s a very important phase, isn’t it? Obviously, this phase will influence all consequent phases. So when I realized that I started looking around on how to do it. I checked various research papers, OpenCV functions, tutorials.After doing so and trying some methods (check my GitHub for the code snippet that I tested) like image thresholding (classifying pixels into two groups: below the threshold or above the threshold), dilation and erosion, contour approximation, MSER regions detections using Neuman proposed method, I came to the conclusion: to improve text detection and recognition it needs much of parameters understanding and tuning for these methods  whereas I’m extremely limited by deadlines imposed by my client!

Wow, how funny is that! What should I do! HEEELP! Deadlines and newbie in computer vision!… Yeah, I felt frustrated! You surely faced such a situation in a specific phase of a project.

The Me surprised about how much I have to learn and deal with.

Fortunately, when remembering my motivation and goal I decided to keep calm and raise the challenge 2 in my project path! So I read some other papers with more attention to the details of the work presented and I got to a conclusion!

Oh, finally we are in the conclusion. What is it? And still, I don’t get what you mean by that “new dimension”!

First, the images I’m dealing with have some specific characteristics:
– low contrast between the text and background,
– The distance between characters presents variations,
– The number of characters is not the same everywhere,
-And the text doesn’t involve words, it means it can’t be recognized using a dictionary and a list of choices, but a mix of numeric characters and Alphabetical characters.

With consideration to that, I examined papers with methods based on Convolutional Neural Network and decided to go for it because according to research and some applications of CNN it delivered much better result! Especially Text-Attentional Convolutional Neural Networks for Scene Text Detection method is what interested me because it is based on a more informative supervised information and on improving contrast in order to detect text.

mmm.. more informative supervised information? Interesting..How is that ensured?

The presented method involves:
-training a CNN with more informative supervised information, such as text region mask, character label, and binary text/non-text information.
-introducing a deep multi-task learning mechanism to learn the Text-CNN efficiently and making it possible to learn more discriminative text features.
-developing a new type of MSERs method by enlarging the local contrast between text and background regions.

(If you know a better paper method that seems to fit my problem, let me know 😉 )

And since that time, which was about approximately two weeks ago, I have started learning Deep Learning with computer vision! That new dimension in machine learning! I tried to familiarize myself with the concept in a fast way so I read these articles that I suggest to you if you want to start with deep learning and computer vision:

A quick introduction to Neural NetworkAn intuitive explanation of Convolutional Neural Network, and A beginner’s guide to understanding Convolutional Neural Network. After that, I installed Tensorflow CPU as it’s easier to install than GPU version and Keras because it’s easier (more high level) than Tensorflow and followed Machine Learning Mastery website tutorial: develop your first Neural Network in Python with Keras step by step. There, I discovered something very important! I really need to switch to GPU version because it’s very slow! Buuuuut… to run GPU version, you have to use the Graphics card (NVIDIA) which I found out later that it doesn’t work on a Virtual Machine!!! Oh my God! A new time-consuming thing with an approaching deadline 😀

Apart from that, I didn’t mention that I have to resolve a very important problem too, that was the first reason of trying out other methods before switching to CNN: I don’t have a big amount of data! (you may tell me how then you will make it work with deep learning! Everyone knows that it’s the most important component of Neural Networks! Otherwise, you can get good results because simply your CNN won’t learn much enough from your input images!)…

It’s possible! There are some methods, some of them are less evident than others, but it’s possible!

I will explore more this 3rd challenge later one but for now, I’m very happy to find a complete article about Setting Up Ubuntu + Keras + GPU for Deep Learning which was published yesterday by PyImageSearch! (Or maybe there is another way…Do you have in mind another possible way to get a CNN running faster? )

By the way, if you are starting like me in Deep Learning with computer vision I would be happy to share with each other the feedback of what we experiment along the way because that will help us move faster ;).

And if you are experienced in the field, let me know what you think about my taken strategy in dealing with the project and what possible methods that can help.

Up for your comments :))

 

 

 

 

Applied text classification on Email Spam Filtering [part 1]

Applied text classification on Email Spam Filtering [part 1]

My activities, projects, Technology

Since last few months, I’ve started working on online Machine Learning Specialization provided by the University of Washington. The first course was about ML foundations, the second about Linear Regression, and the third course on which  I’m currently on track is about Classification. I liked the courses almost in every aspect as they teach ML algorithms implementation from Scratch. That was my goal when I decided to discover the field in more depth. But, honestly, I was feeling that there is a gap somehow because many questions were left without answers along the way. Then, after reading about how to start with Machine Learning, I found that most articles emphasized the importance of combining courses with practical projects in order to apply what’s learned and better assimilate it.. and it’s so true! You just have to try to combine both and you will soon notice the difference!

So, here I am in my first practice application! 😀 I chose to try Email Spam Filtering as it’s a very common topic in applied classification. It can be understood easily because we are experiencing spam filtering in our e-mails every day.

I followed a simple starting tutorial: Email Spam Filtering: A Python implementation with Scikit-learn. Soon after finishing it, my brain started analyzing the steps and a bunch of questions bombarded my head!

Why “equal number of spam and non-spam emails”? What’s stemming? Are there other methods to clean data other than removing stop words and lemmatization? How is the partition between the training set and test set done? Why no validation set? Why Naive Bayes classifier and SVM (Support Vector Machines) were specifically used? What makes Naive Bayes so popular for document classification problem? etc..

As William.S.Burroughs said, “Your mind will answer most questions if you learn to relax and wait for the answer.”

I took a breath and started answering question by question by doing sometimes search on the net, experimenting some changes in code and analyzing the output. And I’m happily sharing  the results:

1) The data we need

– how many emails we’ve seen (will be used in train-test sets)
– how many emails go in each label (used to detect if there is imbalanced data)
– how often a word is associated with each label (used to calculate the probability of an email being a spam or ham (class 0 or class 1))

2) Cleaning data

Why cleaning the words list? Cleaning data is essential in order to reduce the probability of getting wrong results because some words have no influence on the classification (they can neither be associated with spam class and nor with ham class) and there are words that can be normalized in order to group same-meaning words and reduce redundancy. By acting on the quality of the training data, we can change what is called the accuracy of the classifier. So removing stop words, stemming, and lemmatization help in improving the results of Machine Learning algorithms.

3) Naive Bayes

Why was Naive Bayes used? Naive Bayes has highly effective learning and prediction, it’s often used to compare with more sophisticated methods because it’s fast and highly scalable (works well with high-dimensional data) and as Andrew Ng suggests when dealing with an ML problem start by trying with a simple quick and dirty algorithm and then expand from that point.

How is Naive Bayes simple and easy? Naive Bayes is based on “Bayes” theorem and it’s called “Naive” because it assumes that features are independent of each other given the class(no/little correlation between features), which is not realistic. Thus, Naive Bayes can learn individual features importance but can’t determine the relationship among features. Besides, the training time with Naive Bayes is significantly smaller as opposed to alternative methods and it doesn’t require much training data.

Why Multinomial Naive Bayes? What about other models like Gaussian Naive Bayes or Bernoulli Naive Bayes?

Well, Multinomial NB considers the frequency count (occurrences) of the features (words in our case) while Bernoulli NB cares only about the presence or absence of a particular feature (word) in the document. The latter is adequate for features that are binary-valued (Bernoulli, boolean).  Whereas, with Gaussian NB, features are real-valued or continuous and their distribution is Gaussian, the Iris Flower dataset is an example with continuous features.

4) Support Vector Machines (SVM)

Why was SVM used? I didn’t find a specific reason to that, but what I learned is that SVM delivers high accuracy results because it uses an optimization procedure. SVM builds a classifier by searching for a separating hyperplane (optimal hyperplane) which is optimal and maximises the margin that separates the categories (in our case spam and ham). Thus, SVM has the advantage of robustness in general and effectiveness when the number of dimensions is greater than the number of samples.

Unlike Naive Bayes, SVM is a non-probabilistic algorithm.

What’s the difference between LinearSVC and SVC (Scikit-learn)? The difference is that they don’t accept the same parameters. For example, LinearSVC does not accept kernel parameter as it’s supposed linear. SVC supports more parameters(C, gamma,..) since it holds all possible kernel functions (linear, polynomial, rbf or radial basis function, sigmoid).

How can tuning SVM parameters be done? Tuning SVM parameters improve the performance of the algorithm. Some of them have a higher impact:

-Kernel: Kernal is like a similarity function. It’s a way of computing the dot product of two vectors in possibly a high dimensional feature space using data transformations based on some provided constraints into a more complex space. Kernel functions are sometimes called “generalized dot product”.

-Gamma: Kernel coefficient for ‘rbf’, ‘poly’ and ‘sigmoid’. Higher the value of gamma will try to exactly fit as per training data set i.e. generalization error and cause an over-fitting problem.

-C: The factor C in (3.15) is a parameter that allows one to trade off training error vs. model complexity. A small value for C will increase the number of training errors, while a large C will lead to a behavior similar to that of a hard-margin SVM.” Joachims (2002), page 40

5) Analyzing output in different cases

What if I vary dictionary size?

Varying dictionary size means changing the number of features (words). So, I wanted to explore the impact of having more features and what’s the limit of a good result based on confusion matrix result.

I tested on size= {3000,5000,6000,7000} and discovered that at size = 7000, SVM classification starts slightly dropping (false identification) while Naive Bayes delivered same results despite the size variation.

I think that at that point maybe target classes started overlapping or training data overfitting. I’m not yet sure about the explanation of the result.

What if I try Gaussian and Bernoulli?

Obviously, introducing Bernoulli won’t help because as I explained above, it doesn’t provide enough information in our case, we need the count of words, not the presence/absence of it.

Multinomial NB:
[[129   1]
[  9 121]]
Gaussian NB:
[[129   1]
[ 11 119]]
Bernoulli NB:
[[130   0]
[ 53  77]]

As we can see, Multinomial NB outperformed both Gaussian NB and Bernoulli NB.

What if I try GridSearch on SVM in order to tune SVM parameters?
Params GridSearch: param_grid = {‘C’:[0.1,1,10,100,1000],’gamma’:[1,0.1,0.01,0.001,0.0001]}
Best params found: Gamma: 0.0001 ; C: 100

Linear SVM:
[[126   4]
[  5 125]]
Multinomial NB:
[[129   1]
[  9 121]]
SVM:
[[129   1]
[ 62  68]]
GridSearch on SVM:
[[126   4]
[  2 128]]

SVM tuned parameters using GridSearch allowed to get better results.

Conclusion

So, that was about my first steps in Email-Spam Filtering application! I hope that it’s helpful if you are thinking about starting a text-classification project! And I will continue sharing various reflections and experiments along the way. For next time, I will explore more the improvement/change of training data and features.

The project is on Github too.

Tell me about your experience with text-classification? In what did you apply it? What methods do you suggest to apply? What are the challenges?

Helpful Resources:

[1] Naive Bayes and Text Classification.
[2]Naive Bayes by Example.
[3] Andrew Ng explanation of Naive Bayes video 1 and video 2
[4] Please explain SVM like I am 5 years old.
[5] Understanding Support Vector Machines from examples.

Starting with OpenCV and Tesseract OCR on visual studio 2017 [Challenge 1]

Starting with OpenCV and Tesseract OCR on visual studio 2017 [Challenge 1]

challenges, Computer Vision, learning, projects

I have recently started working on a Freelance project where I need to use text scene recognition based on OpenCV and Tesseract as libraries. I was so motivated to hit the Wolrd of computer vision combined with machine learning and experience developing applications in the field, so I welcomed challenges that come with!
Here I’ll be talking about the first challenge and how I tackled it.

found myself with multiple new things to prepare in order to start coding, without mentioning that it’s been a long time before when I last coded with C++ (back to my university time)!

At first, I was asked to use OpenCV 3.0 and Tesseract 3.02 in order to run the project’s part which is already available. So I installed OpenCV 3.0 and Tesseract 3.02 with Leptonica Library, by following provided documentation about how to build application with OpenCV on Windows Visual Studio in this link. Then, I tried to run the project in Visual Studio 2017. I got more than 800 errors!!! Most of them where LINK errors of type:

mismatch detected for '_MSC_VER': value '1700' doesn't match value '1900' in <filename.obj> 

and

error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MT_StaticDebug' doesn't match value 'MD_DynamicDebug' in file.obj. 

The first error was caused by the fact that objects were compiled by different versions of the compiler which is not supported because different versions of the standard library are binary incompatible. Actually, the numbers 1700 and 1900 meant binary version of  v11 (Visual Studio 2012) is not compatible with the binary version of  v14 (Visual Studio 2015).[1][2]

About the second one, the reason behind is that both the library and your project must be linked with the same settings with regards to the C Runtime Library, and that whether in Release or Debug mode [3]. And here the solution in case you need to use static libs or you can deactivate static lib when building with CMake by setting:

-DBUILD_WITH_STATIC_CRT=OFF

 

So I decided to change the version of Tesseract so that it will be compatible with newer version of Visual Studio and x64 target project. As compiling Tesseract3.05.1 with cppan client was not possible for me (ccpan crashes just after launching it), I decided to go for VS2015_Tesseract. I followed these steps in addition to using v140 toolset because Visual Studio 2017 was shipped with v141 toolset, but it supports v140.

Besides, there were some other points to consider. Unfortunately, I got again errors after configuring my project exactly like it’s in tesseract\tesseract.vcxproj (extracted from VS2015_Tesseract) and I find out that the reason behind was: Having a “space” in tesseract path directory.

One last thing to consider when using this tesseract version is to the initialization of tesseract: TESSDATA_PREFIX environment variable is required to setup Tesseract data (tesseract/tessdata) folder. Also you can set it via api->Init(<data path here>, <language>)

I wanted to get also a newer version of OpenCV and went for OpenCV 3.02. Fortunately, that was easier than Tesseract installation, but of course not without some difficulties. I followed this Tutorial in order to install it. In these steps, you have to pay attention to the version of visual studio you choose when configuring CMake as there is a version for win64 (or x64) target project and a version for win32 (or x86). I chose “Visual Studio 15 2017 Win64”.

And yes! it worked! I tested opencv3.2 on a sample project and tesseract on a sample project with OpenCV! I was relieved! Finally, I could start coding!…

After launching the project “End to end text recognition“..I discovered that I wasn’t so close! There was a problem with opencv_text module. I got the error: Tesseract not found at runtime. So, I asked for help through OpenCV forum and StackOverflow and got a suggested solution. In the step of OpenCV configuration using CMake, besides providing Lept_LIBRARY Path and Tesseract_INCLUDE_DIR Path and tesseract_LIBRARY, check Build_examples checkbox, uncheck Build_tests and Build_perf_tests (You don’t need tests when you are a beginner and it takes a lot of memory space, as I know). Also, be sure that:

– tesseract305.dll file is in tesseract-3.05.01\build\bin\Release directory;

leptonica-1.74.4.dll file in leptonica-1.74.4/build/bin/Release.

-put them both in opencv\build\bin\Release (if you will use Release mode or bin\Debug if you will use debug mode).

After that, you have to open INSTALL.vcxproj from OpenCV build directory and build it in Debug (if you will need Debug mode later) and then Release. Be sure to select the right mode x64 (if you chose x64 in CMake configure) or x86 (if you chose Win32 in CMake configure), or you can get errors of type:

“fatal error LNK1112: module machine type ‘x64’ conflicts with target machine type ‘X86”  [4]

That’s all! Now you can start creating projects in Visual studio. Don’t forget to configure the project correctly:
C/C++ ->Additional Include Directories and Linker->General->Additional Library as Well as Linker->Input->Additional Dependencies in a similar way that the one I did when configuring Tesseract.

To test one of opencv text module sample example, put tesseract.dll and lept.dll in the same directory as your .exe (opencv/build\bin\Release\<projectname.exe>) and copy all files from opencv\opencv_contrib-3.2.0\modules\text\samples to opencv/build\bin\Release\

Note: for me, I didn’t test creating a project because when updating Visual Studio 2017, I had a bug!!! Which I tried to resolve and at the end I switched to Ubuntu 16.4!! 😀 .. and how easy it was because these libraries have a greater community in Ubuntu and there is video series to install latest versions. You just need to follow this Francesco Piscani youtube channel videos on how to install OpenCV, Tesseract and Leptonica. Also, you can choose the option to download Tesseract3.04 and Leptonica1.73, which are older versions that don’t require build from source. But I don’t know how to run the project on Debug mode and if it requires another build condition of opencv library. (like in Windows you have to choose build configuration Release or Debug)

What about your experience with OpenCV, Tesseract and Leptonica? Are you working on any project involving Text recognition in natural images? Share in a comment!

Please don’t hesitate to write me in a comment if you tried the steps and failed! I will do my best to help! 🙂

I did it! Yes! Yes! I've passed through struggle moments, didn't move from my laptop for almost all past 4 days.. sometimes it was hard to keep trying.. thought about giving up multiple times but couldn't accept that at last!.. Went through Stackoverflow, opencv forum, github, visual studio 2017 buggy version 15.3.. 😥 And finally here I'm.. impressed by the simplicity of Ubuntu, which I switched on thanks to my husband 👫 who convinced me to use it in order to make the project work 🙌 ❤. He showed me a working example of opencv, just to support me. RESULT: End to End text recognition sample example of opencv with tesseract-ocr is working! Now.. some magic of machine learning needs to be there ✌🎉 BTW: I will write in the coming 2 days a summary of this challenge of starting with opencv and ocr detection 😉 so that you don't loose much time on looking on ways to do so if you think about starting on it. #computervision #perseverance #machinelearning #womenintech #womenwhcode #datascience #ocrdetection #c++ #opencv #photoOCR

A post shared by sarah mestiri (@sarahmestiri) on

[1]: Mismatch LINK Error explanation
[2]: MSVC internal numbers explained in Wikipedia
[3]: Static VS Dynamic Library Linking
[4] Explanation of different target machine types problem.

Other LINK Errors

J On the Beach 2017

J On the Beach 2017

My activities, Technology

This time I made it to Malaga, Spain!

I attended J on the Beach 2017 conference, a Big Data event targeting developers and DevOps communities. It was held over 3 days.

The first day was about workshops. In the morning session, I participated in “Building Microservices on DS/OS” ensured by Jorg  Schad where I got an idea about containers, Mesos, Clusters. It was a little bit difficult for me as all of them were new.


The Afternoon session was  “Hands on Elastic Stack 5.3”. The presenters Pablo Musa and David Pilato were very helpful and I liked so much combining discovering these tools with practice, as I believe that the best way to motivate people about learning something, is by showing results!

I discovered how it’s possible with Elasticsearch to get results of a search query over very large dataset in milliseconds. Elasticsearch has unique capabilities thanks to its indexation process.  Not only can you process a query rapidly, but also you can visualize and explore your query results easily with Kibana! Those where the parts, I most enjoyed, apart from Logstash and Filebeat.

The second day was the conference opening day and it started BIG! Eric Ladizinsky presented to us the evolution of Quantum Computers.

Eric justified weak machine learning binary classification and how quantum can make machine truly intelligent by enabling Probabilistic Machine Learning. More about Quantum age can be found in D-Wave company website.

Various sessions were held after that. The schedule was partitioned by field of interest and level to ensure there is food for everyone’s mind.

As my interest in machine learning couldn’t take down the one for Java, I was excited to attend “Real World Java 9” talk, by Trisha Gee. And I wasn’t the only one! The room was full!

Trisha started by mentioning that Java 9 is still not stable but that it’s a very promising version with its new features! She did her presentation with Live coding (Oh, Yesss!) where she tried the new Flow API and Streams API. More about her presentation here.

If we speak about BIG Data, then for sure there is a subject not to miss: Performance.

Rob Harrop presented a session about “designing and running performance experiments”. It was an interesting session! Rob emphasized that wait times are A THING. Wait times are drawn from a distribution and if you don’t know the distribution, exponential is a good fallback.

James Allerton spoke about how to develop, deploy and iterate often using Oracle management API. And as concretization is always a better way to convince, he demonstrated with a live demo!

Day one last session was about “Dynamic data visualization”. Santiago Ortiz created the link between human intelligent learning capabilities and machine learning created models (classification, labeling, prediction). Trying out things with his children, like classification, made him conclude that understanding human perception and cognition is instrumental in creating environments in which models, machines, and humans collaborate.

Before heading to day two, I like to mention one other thing: Oracle IOT digital demo. Fascinating! Isn’t it?

I couldn’t resume the first day with enough words, it was such a great one! Especially, when it ended with something related to women in tech (obviously, as I’m proudly one of them 😀 )

Carmel Hassan presented to us the visualization of collected data from JOTB17 day one in a way that showcases diversity facts. Surprisingly, the number of women attendees decreased between last year and this year from 13% to approximately 10%! While women are not well represented “live”, they tend to have greater percent “online” according to the data collected from the interaction with JOTB17 publications. Well, it seems that there is still a lot of work to be done in the context of women in tech! But no worry, superwomen heroes are there! Check Yes We Tech community if you want to support!

Here we are! Now, day 3.

To me, what marked the conference day 2 most is “Distributed Sagas” session by Caitie McCaffrey.  Distributed Saga is a collection of requests and compensating requests that represent a single business level action. It makes building microservices easier because it allows putting different services like for example renting a car, booking a hotel reservation and booking a flight ticket all together before proceeding to payment action (one payment action for the three of them). Caitie presentation really added value to my knowledge on distributed services. Her presentation can be found here.

Otherwise, for Elasticsearch users, here is an advice from Pablo Musa: Do not Overshard (Don’t just keep default settings).

So, that was about J on the beach event! But was it all? Where is the fun part?

Well, I didn’t want to mention all the fun at first, otherwise, tears will take place for those who missed it! 😀

We had a lot of fun! Spanish Omlette, Paella, JOTB Afterparty, nice organizing people and Malaga who welcomed us with its beautiful weather!

I would like to finish this article with something that I cherish much. I googled it in order to find a quote about and I think that this one describes it well:

The richest people in the world look for and build Networks. Everyone else looks for work.” Robert Kiyosaki

The World may seem big for some people, but not for me! The World is little and magical! I got to meet new friends from Ireland, Spain, Russia, Italy! We discussed many topics together, we discovered each other countries culture! We ate ice cream :D. (One of my friends will recognize himself :p by the why of ice cream) and Yeah, of course, laughed a lot!

Thank you all and see you next time!

Note: You can find more about other sessions by using #JOTB17 tag on Twitter!

A glimpse into AI potential between today and tomorrow (part 1)

A glimpse into AI potential between today and tomorrow (part 1)

My activities, Technology

April 23rd, I was honored to be invited by GDG Monastir (Tunisia) as a speaker in Google Next 2017 extended event. My talk was about Artificial intelligence and following is the written version of the idea I wanted to communicate.

I bet that when you think/hear about Artificial Intelligence many thoughts pop to your mind! Big Data, IBM Watson, Alexa, Google self-driving car, NAO, Turing, Google Cloud, but what’s Artificial Intelligence? Is it a brain that can outperform human’s brain? How does AI impact our everyday life? What are the technologies behind AI?

What’s AI?

Back to 1950, when Alan Turing, a computer science pioneer, started thinking about “What if machines learn?”. He invented Turing Test in the aim of making an intelligent machine. The test was an adaptation of a Victorian-style competition called the imitation game. At that time, Alan pinpointed a possible horizon of AI that many didn’t believe on or ever imagined.

It’s only 5 years later that, in Dartmouth Conference, John McCarthy coined the term of “Artificial Intelligence”, defining it as “the science and engineering of making intelligent machines”.

Intelligent machines, artificial intelligence.. well let’s define intelligence before digging deeper.

What Intelligence consists of [1]
As we can see, Intelligence consists of:

  1.  the ability to process data,
  2. learn,
  3. solve a complex problem in special purpose as well as general-purpose domain,
  4. ability to reason and to draw inferences based on the situation.

Do the machines have all these abilities? Some of you will say “I think yes!” others will object..and what I suggest here is to discover together the answer by checking where is artificial intelligence now! because there is no better convincing argument than real facts!

What’s the impact of AI?

AI has an impact in different industries: automobile, e-commerce, marketing, robotics, mobile and the list is still longer! And one of the most important pillars of a human being is, without any doubt: health. No one could make any advancement in technology if his health situation was deficient somehow. So, investing on AI in the health sector is a big deal for many companies including IBM and Google.

Artificial Intelligence is helping doctors better explore the human body and deal with its complexity. April 2017, a research conducted by a team at the University of Nottingham in the United Kingdom, proved that doctors can better predict patient’s heart attack risk when using AI. This can save thousands or millions of people every year.

The fact is each year, between 15 million and 20 million people fall victim to cardiovascular disease, which includes heart attacks, strokes, and blocked arteries. Doctors use a method called ACC/AHA, based on some guidelines like age, cholesterol level, and blood pressure to predict the risk of heart attack. Actually, the method is revealed to be not very efficient. The complexity of human body still hides some unknown biological guidelines.

Fortunately, as Stephen Weng, an epidemiologist at the University of Nottingham, told Science Magazine. “What computer science allows us to do is to explore those associations.”  Trying out some machine learning algorithms like logistic regression and neural network on the problem, resulted in better predictions! The algorithm was 7.6 percent more often than the ACC/AHA method and resulted in 1.6 percent fewer false positives. That means that in a sample size of around 83,000 patient records, 355 additional lives could have been saved. [2]

Achievements of AI in medicine don’t end there:

  1. IBM Watson for oncology is working cancer diagnosis,
  2. Google DeepMind blindness detection: Two million people are living with sight loss in the UK, of whom around 360,000 are registered as blind or partially sighted. [3] The research project is investigating how technology could help to better analyze scans, giving doctors a better understanding of eye disease.  The team working on developed an AI algorithm that can automatically identify diabetic retinopathy, a leading cause blindness among adults. “We were able to take something core to Google—classifying cats and dogs and faces—and apply it to another sort of problem,” says Lily Peng, the physician and biomedical engineer who oversees the project at Google.
  3. Robots as surgery and patient assistants like Romeo, developed by SoftBank for elderly assistance.

While it’s true that a lot of work still need to be done in order to actually help developing medical services and patient assistance, AI impact is already there and doctors believe in how collaboration with AI will be beneficial, “I can’t stress enough how important it is,” Elsie Ross, a vascular surgeon at Stanford University in Palo Alto, California, told Science, “and how much I really hope that doctors start to embrace the use of artificial intelligence to assist us in care of patients.”

Moving on to the first field that grew with AI: Games.
The story started with computers beating humans at games like chess or Jeopardy by simply calculating all possible moves on a board or rote learning.

Today, computers may even have the ability to think ahead and reason! Researchers were not considering these possibilities before Google’s Alpha Go win!

March 2016, Lee Se-dol, one of the world’s top Go players, won just one of the matches against the AlphaGo program. The game is considered to be much more complicated than chess, as it has much more move possibilities ( > 10170 legal positions).  Thus, it usually requires a special human skill: intuition and creativity.

DeepMind researchers said that kind of computing power has real-world promise outside of gaming, especially in health care and science.

Health, Gaming, what else?

Google self-driving car

Well, people are now talking about vehicles that not only transport us but also think for us!

Google began testing a self-driving car in 2012, and since then, the U.S. Department of Transportation has released definitions of different levels of automation, with Google’s car classified as the first level down from full automation. Other transportation methods are closer to full automation, such as buses and trains.

AI application areas in transportation field are quite diverse. Thanks to pattern recognition, system identification, classification and many other subsets of Artificial intelligence, transportation has known a huge evolution over the years!

Traffic state tracking, advanced driver assistance systems (ADAS), driver alert systems (e.g drowsiness detection) already exist to provide drivers with a better driving experience. Google, Tesla, Volvo as well as many other companies are focusing on providing road safety in a way that “Artificial intelligence could change our emotional as well as our practical relationship with cars“.

Ok, you may tell me: that all sounds good but I just don’t have all these applications of AI around me! “I’m not in the USA to see self-driving cars”, “I’m not a doctor to try out AI assistance in my job”, “I don’t play games!” I’m not sure about the impact of AI in my life!

You are already using AI! It’s not a science fiction! It’s real. Maybe, you are just not aware of it! And whether you want it or not, Artificial intelligence is influencing each aspect of your everyday life.

While surfing on the internet you are experimenting user behavior tracking. Many companies like Google, Amazon, Facebook and others are using your preferences, history, searches, and purchases, in order to predict your interests and target you according to the situation (Facebook news feed, ads on the right/left side of a website page, suggestions of books to buy,..). So, through Google search, Facebook, movies recommendations in Netflix, Amazon store, you encounter aspects of Artificial intelligence.

AI is within your mobile too! Siri, Google Now, Cortana are all intelligent assistants using AI technologies designed to help you get answers to your questions, or perform actions with voice control. For example, you can say: “What’s the weather today?” and the assistant will search for the information and relate it to you.

Not only that, AI is also used for online customer support. Not every website you visit has a live person communicating with you. In many cases, you’re talking to a rudimentary AI (chatbots).

You see? AI is everywhere! Your smartphone, your car, your bank and your house possibly too!

 

At this point, the story is starting to be like a horror film: talking about all these AI impacts is raising some questions..it seems controversial, right? Intuition, creativity, assistance, holding your data and maybe risking your privacy, using data to train machines, helping doctors, distracted driver-spotting, Humanoid robots, what’s next? Does Artificial intelligence hold real “Intelligence” (that we defined at the beginning)? Is AI going to take our jobs? 

To be continued…

Resources:

[1] Artificial intelligence defined, Deloitte.

[2] Can machine-learning improve cardiovascular risk prediction using routine clinical data?

[3] Google DeepMind research in health.

Is cracking the coding interview the only benefit of learning algorithms?

Is cracking the coding interview the only benefit of learning algorithms?

challenges, learning, My activities, Technology

Often, algorithms are considered only when someone is looking for a new job. This tight perception of algorithms use puts us away from what algorithms can allow us to achieve!

Actually, algorithms are everywhere! Algorithms are involved in each aspect of computer science! Not only that but also used in a wide range of fields: recommendations, social media, medicine, psychology, transportation and the list is longer still!

Anything you do, can be broken down into small steps and that, is an Algorithm. Imagine you wake up the morning to go to work and you can’t remember where are your car keys, how would you find them? One approach might be to apply an algorithm, which is a step by step logical procedure. First, you look for places where you used to put it in. You try to remember the last time you used it. You check the place where you went when entering the home. Sooner or later by the flow of steps, you eventually find the car keys. An algorithm is a sequence of instructions which ensure a certain task completion. We, as human, apply algorithms in order to take actions in every aspect of our lives.

An example of how to prepare a Smoothie. It needs an input which is Fluid Frozen fruit Flavorings, and a sequence of ordered steps to be executed to deliver a Smoothie as an output. These are algorithm components. (original resource of the picture)

So, algorithm knowledge can be an asset in improving your life or even others’lives! You can turn something that takes the time to decide on it into an algorithm that assists you in the decision process. You can create an algorithm in order to prioritize your tasks, instead of doing it by a simple to-do list. You can make an algorithm in order to predict the best time to visit a city. You can make algorithm in order to get suggestions on what films to watch according to your preferences and historical data. There is no limit to what you imagine and can make real by applying algorithms!

Now, I think it’s time to start talking about how to gain those set of skills in order to make a difference with algorithms! First, let’s go for the characteristics of an algorithm.

What are the characteristics of an algorithm?

  1. It should be finite: If your algorithm never ends trying to solve the problem it was designed to solve then it is useless
  2. It should have well-defined instructions: Each step of the algorithm has to be precisely defined; the instructions should be unambiguously specified for each case.
  3. It should be effective: The algorithm should solve the problem it was designed to solve. And it should be possible to demonstrate that the algorithm converges with just a paper and pencil.

What are the steps to learn algorithm coding?

  1. Develop your programming logic. Here some ways to improve it.
  2. Pick up a programming language you are comfortable in (if you don’t have any, I suggest learning python, it’s easy and simple)
  3. Learn Data structures: Start with the basic ones: String, Vectors, Lists, Arrays, Map. Then Queue and Stack. Finish with Advanced ones: trees, graphs, tries.
  4. Practice coding data structures. Here is a good link
  5. Learn simple algorithms first and then move to the most common algorithms. Make sure you understand every step and don’t be frustrated from being slow. Remember « It does not matter how slowly you go as long as yo don’t stop ». You can look for videos because they make it easier to assimilate.
  6. Start tackling problems (e.g in Hackerrank, there are funny and challenging algorithms). You must have a strategy of resolving problems, I’d suggest following the method proposed by Gayle Laakmann McDowell.

Finally, through my own modest experience in problem-solving, here are my pieces of advice :

  • Ask questions: « why » and « what » are really important in order to understand the basics of anything! When you question the choice, you will better memorize the answer and you will dig deeper into the source of the information.
  • To stay motivated and keep the learning curve up, create a simple daily useful algorithm after getting the basics. Accomplishment fuels your energy.
  • Try to solve the problem manually by yourself first. Take a paper and start writing steps. Especially take your time and don’t move on to the solution fast.
  • Even if you figure out the solution, don’t tell yourself « It’s ok now, no need to implement it, I know how it works » It’s a big trick there because you never know what you missed if you don’t try it and test it with corner cases.
  • Be patient. If you begin to feel frustrated about a problem, leave it and go do something that is not so stressful. Then, come back.

I hope this helps you to leverage the power of your brain, by discovering its super capabilities to execute algorithms and learning how to translate them into code! 😉

Remember how was Rome build (not in a day right)? By each small step in learning algorithms, you are growing your potential to contribute to making life better in this world!

Suggested books:

Github resources:

mission-peace

Online resources:

Topcoder: an explanation of data structures and algorithms.

ideserve: contains a visual explanation of some algorithm problems

Youtube resources:

Tushar Roy : explanation of some advanced algorithms

Up to my new tech challenges!

Up to my new tech challenges!

challenges, Technology

You have certainly heard about that unique thing, which changes your life and makes it magic, beautiful. It can be a person, a situation, a hardship, a moment of discovery.. so called “The Spark“!

It all started three or four years ago when I discovered those intelligent features in my Smartphone  “Samsung Galaxy S3”: social tag,  smart stay, smart alert and S voice.

Many questions were coming to my mind about those features, I was so curious about what’s behind their functionalities and how humans succeeded in developing such intelligent technologies! At that time, I was studying computer science engineering and wondering which option modules to take for the last semesters. So, I decided to study “decision making and artificial intelligent (AI) systems” as an option for my last two semesters at university. Through this option, I would dig deeper in machine learning and find an answer to my questions.

Unfortunately, the modules were full of theory. As I’m a practical person, I lost my motivation and was convinced that it’s not the right field for me. One year later, having started my work as a software engineer,  I began to feel that monotony of repeating same things every day, without a space for initiative and innovation. Although, it was an opportunity for me to work for a big name company and discover SCRUM methodology as well as learn who to collaborate with my team and build a clean code following high convention standard, I felt something is missing. It was the right time to better shape my future and re-think the direction I pursue! Passing through that tunnel needs a lot of reflection and a clear vision of your purpose and mission!

images-2-2-300x147

Determined to draw my own success path and make the difference, I completed OCJP Certificate, TOEIC Certificate, and started looking for a better job offer where my energy is fueled and where I gain more knowledge in recent technologies and I got it! 

One year later, I’m here again for a new challenge! I couldn’t resist to the World of Data and Analytics as well as the need for machine learning in order to ensure better use of data! So, I attended Data Natives 2016 in Berlin which was an event covering tech trends and innovations in big data and machine learning. By the way, you can find more details about my trip here. From there, a new horizon has just come!

My Network began expanding through Twitter. Everyday, I find myself more connected to the world I love! Hundreds of interesting articles about Artificial intelligence, machine learning, big data and analytics! New people around me sharing with me my passion! 

And it doesn’t stop there, among all the subjects, especially, two subjects caught my attention : recommendation systems and marketing analytics (ads bidding, extracting informations from ads, etc)! How fascinating they are! That’s why, I decided to follow machine learning specialization in Coursera, learn Python and start doing some sample projects either in kaggle platform or by exploring github or just trying to make a project from an idea I have! So, I would welcome from you any suggestions on a project or a resource that can be useful, or maybe joining me in this journey!

I know it’s not going to be as simple as it seems but I’m convinced that I can do it! I’m confident in my ability to learn quickly because my drive and my passion fuel me!

On the other side, I’m willing to keep my skills in Java updated and continue forging them. I will keep that part for another article!

For sure, no new discoveries come along without questions! As I’m digging deeper in machine learning, there is a question crossing my mind: why most of those working with machine learning and data analytics don’t use Java for it?Is it because of the libraries that the other languages like Python or R offer?I really like this language even though it’s more complicated than python (but maybe that’s what makes me like it more 😀 )! Does anyone have an answer to that? Feel free to comment!

Finally, my favorite conclusion is to finish up with some quotes that inspire me and guide me through this way!

“The man who moves a mountain begins by carrying away small stones.”  Confucius

“The future belongs to those who believe in the beauty of their dreams” Eleanor Roosevelt

 

Up to the challenges! And yeah! difficult roads often lead to beautiful destinations. 😉

 

Data Natives Berlin 2016 day 2

Data Natives Berlin 2016 day 2

My activities, Technology

Berlin, October 28th, Data Natives day 2. It was about tech trends, including numerous topics: Deep learning, Source code abstracts, analyzing and searching unstructured Medical Data, smart-data based lending, intelligent hiring, and more.

Klaas Bollhoefer, Chief Data scientist at *um, tackled the gap between several departments in a traditional company and took the example of “operational gap” study, made for one of *um clients. He demonstrated how *um approached  Metro Group issue: what does Metro need to do to become a leading analytics-driven omni-channel retailer. Among that, Klaas stated that “every company requires unique capabilities” in order to reach Data Maturity level and achieve digital transformation. For that, a bottom up approach should be done for analytics in order to explore the framing of any company. That means taking care of traditions and use that to build a new feature. It’s a step-by-step Data thinking model based on capabilities, insights & recommended actions derived from what it really needs to think.

20161028_110613

Moving to finance world, where massive data collection is obviously taking the main part of it, what’s all the collected data from clients is used for? What’s the benefit for the client?! Here came, Patrick Koeck, COO of CreamFinance presentation about “Smart Data-based lending”. Patrick focused on the difference that makes the use of Big Data VS Smart Data. He showcased how smart assessment of data by contextualizing it conducts to a better client experience and a digital transformation. Based on consistency and higher stability with less noisy informations and  more transparent models, smart data helps customer segmentation and decreases the stealing risk  of sensitive data.

img_0141

From data collection and usage comes the struggle to create a value! Martin Loetzsch, CDO of project A ventures, invoked the necessity of creating value as a data scientist in early stage startups by making the link between tech team and business team. He demonstrated the low values creation such as being unclear and for example just ask developers to do something great without taking care of their needs. Thus there is a need to killing black boxes in order to achieve high value creation because people don’t trust black boxes.

cv2rfafvmaef99y-1

Then, Lisa Winter, Sr. Analyst Executive Compensation and Nicolas Rossetti, Risk analyst at Willis Towers Watson, presented to us how to build the bridge between Skillset and use case to become a data scientist. This involves: subject matter expertise, the statistics extension to subject matter expertise and the technology extension to statistics. They used the example of how to build a personalized health insurance pricing and offering as an illustration. Besides, Lisa took the initiative to present to us her “Faces of data science series” or real life data scientists by presenting a list of women in data science as well as men. By the way, I was happy to be mentioned in her list!

IMG_0147.JPG

Well, the hiring process can also be benefiting from AI and Emotional Analytics, this is what Darja Gutnick, Founder and CEO of 12 Grapes, explained to us by stating 5 principles of intelligent hiring, among them goal commitment, accountability and the focus on results. She affirmed that “Alignment on values is everything” and gave the steps of how to predict the success of new hires through measuring skill complementarity and value alignment, according to 12 Grapes model.  Thus, achieving “good hires”.

IMG_0169.JPG

This is about talks, and do you know? With Data Natives, there is also surprises! The conference weclomed on stage Anish Mohammed, Advisory Board Member at Blockchain Advisory Group, for a cyber security talk! Anish explained the link between smart contracts and cyber security. He showcased the Advanced Persistant Threat life cycle, consisting of: incursion, discovery, capture and exfiltration. Also, he presented security analytics evolution and cyber security future scenarios.

IMG_0164.JPG

IMG_0163.JPG

I couldn’t finish the blog without mentioning the “Big Data in FinTech Panel“! It was an exceptional panel where the evolution of FinTech and its Horizons were discussed with the presence of Dr.Stephan Wachenfeld, MD at Savedo, Christian Rebernik, CTO at N26, Jacqueline Faridani, Associate Director of Quantitative Analytics & Credit Risk at VendorMach as panelists and Devie Mohan, CEO at Burnmark as moderator. During the panel, Stephan affirmed that banking is not done anymore by bankers, Christian harnessed the fact that it’s about creating value and using data for the unique benefit of the customer! Number26 is a concrete illustration of that.

20161028_173317.jpg

Well, it’s not all, there is a lot in that day! It was incredible! I had the pleasure to meet Evelyn Münster, a data visualization addict, who encouraged me to follow my dream and keep it up. Otherwise, I was honored to know at the end of the event Elena Poughia, the founder of data natives and Warren Channel, the operations and community manager at Dataconomy Media as well as Maya Lumbroso, the Business Development manager at Dataconomy Media.

20161028_181952.jpg

I would like to thank all the organizers and volunteers. I will be pleased to join their team for Data Natives Berlin 2017, which will be held on 23th and 24th of November.

Data Natives 2016 was a real success! and here a summary about it:

20161028_181739.jpg

As a closing of the event overview, I’d like to present to you Lisa and Nicolas conclusion:

IMG_0157.JPG

See you soon my dear readers! Hope you enjoyed my overview of Day 1 and Day 2 of Data Natives 2016, Berlin-Germany.

Finally excited to announce a scope! I’m eager to attend the next conference dedicated for big data developers for which Luis Sánchez, Digital Marketing Manager at valo.io invited me! So, my next tech stop will be on the beach in Malaga, Spain, to attend the event hosted by valo.io. I’m very thankful for Luis! It was nice to meet him. See you there!

Note: Data natives day one overview: here.

Visit my twitter account @MestiriSarah for more feedback on Data Natives 2016.

Data Natives Berlin 2016 Day 1

Data Natives Berlin 2016 Day 1

My activities, Technology

My journey in Berlin started with two days enjoying its winter landscape and counting days down to the conference for which I was determined to do the whole trip from my country, Tunisia in order to attend.  Thursday, here it is! Data Natives opened its doors!

I got up that same morning very excited to meet people sharing with me  the passion for technology! I already discussed with some of them on twitter,like Lisa Winter and Evelyn Münster, so I was eager to get to know them.

Looking at the schedule, it was difficult to choose which talks to join since all of them were interesting. There were subjects about Visualizing and Communicating High Dimensional Data, Clustering Data, Big Data Semantics, Connected things, and much more here. Besides, as Artificial Intelligence is a field full of diversity in every component, points of view also vary, and that’s why there was a Panel about “the future of AI and Universal Basic Income”, directed by Prof. Dr Hans Uszkoreit, Scientific Director at DFKI and with Nima Asghari, Senior UAV Applications Expert at Drone Industry Insights, Dr. Abdourahmane Faye, Big Data SME Lea, Michael Bohmeyer, CPO at admineo as panelists. The discussion was mainly about the job opportunities drop down due to robots replacement of many tasks done by Human now and by consequence causing a reflection point: is it really a curse? Should we be optimistic about that fact? Are they killing our jobs? Is the Universal Basic Income a solution in order to ensure living cost for those would-be jobless people?

14876361_655826507910789_1573131831_o

Top talks according to me? Well, I’ve enjoyed starting the day with Kim Nilsson, CEO and Co-Founder of Pivigo where I discovered “what does it take to be a successful Data Scientist”. DS potential consists of Interest + Technology (tech skills in at least one of the most commun tools/languages) + Communication (As Data Science is not easily understandable by non-DS people, it requires communication skills so that the information can be transmitted) + Motivation + Hard WorkAcademia.

IMG_0128.JPG

After that, I was pleased to see an example of how data processing and integration can be a benefit when it’s accessible by everyone! This can be done using Open Data. Christina Schoenfeld, Business Developer at Open Data Soft, presented to us the advantages of harnessing open data by making it easy to publish and to understand in Germany, which can be of a huge innovative potentiel, with impact on society and economy.

IMG_0136.JPG

In the other side, Alexandra Deschamps-Sonsino, Director of Design swarm, shared with us an important reflection time about the internet of things and the use of data, questioning: “When things Speak, are we ready to listen”. Many hidden truths behind the data were showcased.. such as the ones using fitness track systems actually do not loose weight while the others  who’re NOT using them Do!More intriguing, the tools used to track elders in order to help take care of them actually made them more susceptible to be left alone, with miserable company (TV and pets) instead of having their children around them.

Oh, yes! there is Data Hypocrisy!..

Many more interesting points were mentioned by Alexandra and I’ll relate one more: she presented some alternatives like “good night lamp” and Data Brick which encourage people to have control over their own data experience.

14795904_655824607910979_1121551727_o.jpg

I’ll say no more, and I’ll let the last words on this blog post be Alexandra’s last presentation’s advice. See you tomorrow!

“Question what is hiding behind your data. Find out what behaviors are you really enabling, good and bad. Think about what is the alternative to using your product.” Because.. ” Data is a slave to use. Use is a slave to experience. Experience is a slave to culture.”

Note:  Data Natives day 2 is available here.