AI and DS Skills

 View Only

  • 1.  Best part in Data and AI

    Posted Thu November 12, 2020 11:29 AM
    Hi All,

    Which one is the best part of Data and AI

    1) Machine Learning
    2) Neural Network and Deep Learning
    3) Computer Vision
    4) Natural Language Processing

    ------------------------------
    Ratnesh gupta
    Team Lead
    IBM India Pvt Ltd
    Noida
    ------------------------------

    #AIandDSSkills
    #DataandAILearning


  • 2.  RE: Best part in Data and AI

    Posted Fri November 13, 2020 08:56 AM
    Hi Ratnesh,

    NN is a part of ML, and DL is a part of NN. Both computer vision and NLP will probably include at least some ML techniques. Except of that, computer vision also includes computer graphics, sensor technology and signal processing. All of these fields are very broad.

    So, the question is, the best part in what sense? It very much depends on your goal.

    ------------------------------
    Lenka Cizkova
    ------------------------------

    Original Message


  • 3.  RE: Best part in Data and AI

    Posted Fri November 13, 2020 09:19 AM
    Hi Lenka,

    Thanks for your response, I just want to know, which field has more opportunity and growth as all of these fields are very broad .

    I have just completed my course on Data Science and looking for project related to Data Science field that's why I am asking same to choose best one



    ------------------------------
    Ratnesh gupta
    Team Lead
    IBM India Pvt Ltd
    Noida
    ------------------------------

    Original Message


  • 4.  RE: Best part in Data and AI

    Posted Fri November 13, 2020 10:00 AM
    Hi Ratnesh,

    I believe that all these fields will continue to grow quickly, and all of them offer many opportunities. My suggestion is to take into account your personal interests, and possibly any relevance to your current or dream job.

    ML / NN / DL are more general, and therefore, somewhat more theoretical; being a basis for more specific methods / approaches / application libraries etc, they will be applicable in most industries. This is certainly a good starting point if you want to be a sort of generalist, or haven't decided yet.
    Computer vision and NLP, on the other hand, will probably be closer to concrete applications (while taking into account that they again consist of many sub-fields). For image recognition, in general or applied, e.g., in autonomous vehicles incl. drones, production robots, space flight, computer vision is extremely important. NLP will have many applications in communication (both text and speech), think about search engines, knowledge management, sentiment analysis, medical reports, automated translation...

    ------------------------------
    Lenka Cizkova
    ------------------------------

    Original Message


  • 5.  RE: Best part in Data and AI

    Posted Mon November 16, 2020 01:31 AM
    Hi Lenka,

    Thanks for sharing your analysis, Can you please suggest me best way to start these , also share links if you have, from where I can get start practical knowledge and can enhance my theoretical knowledge as well.

    ------------------------------
    Ratnesh gupta
    Team Lead
    IBM India Pvt Ltd
    Noida
    ------------------------------

    Original Message


  • 6.  RE: Best part in Data and AI

    Posted Mon November 16, 2020 12:33 PM
    I will share a list of AI/DS courses I'd like to take at some moment and/or interesting and relevant courses suggested by various people in forums etc; I've taken a couple of these for my career switch (math teacher to developer/data scientist with NLP - while at the beginning I thought NLP is the least probable I'd end up with ;)) and I plan to take more of these later.
    Hopefully, some of these will be useful to you and/or other members of the community.

    Basics:
    • some basis in math can be useful: calculus and linear algebra (vector and matrix operations, differentiation...); more will be needed for computer vision (depending on what precisely you'd do with computer vision) - e.g., integrals and trigonometry.
    • statistics: either a basic course of statistics, or work through your AI/data science course(s) and learn about the terms you come along.
    • programming: Python seems to be the most relevant programming language for data science (it's quite easy to learn and there's a lot of libraries for various DS tasks); there are many possibilities to get into Python, I've got the basics via my data science courses and went deeper via 30 Days of Code and Python practice at HackerRank (free; you can get a Python certificate when you pass a certification test, which is also free). You can also start or refresh with this Python course at Kaggle (free, with a certificate). Some more resources:
      • Learn to Program: The Fundamentals by University of Toronto via Coursera (with Python; free without certificate, certificate = 41 euro)
      • Python tutorial (incl. basic ML tools) by w3schools (free)
      • Python 3 course at SoloLearn (free, with certificate) - I've done this course several years ago and I can only remember that it was very easy to follow, cut into small chunks, easy to do even via a tablet etc in various short breaks I had.

    General:
    • Kaggle has a whole spectrum of short courses on many aspects, starting with Python / pandas / data visualization / intro and intermediate ML, SQL, DL, computer vision up to NLP and Game AI; all of them are free and with a certificate. Kaggle also hosts a lot of data files, various challenges, and, last but not least, user notebooks where you can see how people solve various problems - precious! As a Kaggle user, you'll also get some free GPU and TPU time to train models and/or solve the challenges. I have finished 2 Kaggle courses and started another one (and definitely going to do more), my experience is that they are quite easy to follow, strongly oriented towards application of the tools to solve a practical problem, but on the other hand, there's no strong theoretical basis, so, at moments, you can feel like I'm doing something and have no idea what precisely - but it works somehow. Thus, I definitely suggest taking the Kaggle courses and combining them with some more theoretical courses elsewhere.
    • You can practice your knowledge in various domains at HackerRank (and there are certainly other suitable places, I mention the HackerRank because that's what I have experience with and I like it). There are, for example, tasks in AI with subdomains in Statistics and Machine Learning, NLP, Bot Building etc. HackerRank is free.
    • FreeCodeCamp has a lot of free YouTube-based courses; here's an overview of their machine learning courses.

    Intro / starting point / overview:
    • IBM Data Science Professional Certificate (9 courses, the last one is a capstone project, so, when finished, you'll have a project to include in your portfolio), IBM via Coursera (free audit per course without certificate, or monthly subscription; currently, the 1st month is free via this IBM offer). Beginner level (you'll only need some basic math/stats that you can possibly learn while working through this series). I've done this professional certificate; studying full-time, I was able to go through the first 8 courses in about 5 weeks, then I took a break to get deeper into Python (via HackerRank, see above), and finally, came back to finish the capstone. There's some variation in the quality of the courses, some parts are explained to the smallest details while in other parts, you need to take a couple of steps yourselves; some materials would need an update to reflect the updated tools (the may be updated by now, I haven't checked), so, you'd need some patience and determination to find your way (forum participation can be very helpful!) - but after all, combining facts, and finding your way to solve a problem is precisely what a data scientist needs ;) To me, this series was precisely the starting point I needed, so, 5 stars for me.
    • IBM Data Science Professional Certificate, IBM via edX, similar to the professional-certificate series on Coursera but instead of monthly subscription on Coursera, at edX you pay a fixed amount for the whole series (which is 312 euro at this moment) - or you can audit each course separately, which is free. There's another difference: the capstone-project's theme is given at edX, while at Coursera, you may do a project according to your own interests (with some ideas offered for those with a lack of inspiration).
    • Intro to Machine Learning at Kaggle (free); possibly followed by other Kaggle courses.
    • Elements of AI, University of Helsinki (free)
    • Professional Certificate in Applied AI (6 courses, from basic concepts via building a chatbot to computer vision; IBM Watson based) by IBM via edX (free audit per course available; the complete series incl. certificate for 330 euro).

      Deeper into ML/DL/AI/reinforcement learning:
      • Intro to Deep Learning at Kaggle (free; DL and TensorFlow; after Intro to ML; prepares, e.g., for the Computer Vision course)
      • Intro to Game AI and Reinforcement Learning at Kaggle (free)
      • Machine Learning by Stanford University via Coursera, a famous course taught by Andrew Ng (free without certificate; or with certificate for 67 euro); Octave/MATLAB-based (i.e., no Python); differs from most other courses in that you don't learn to use existing tools but construct them from scratch, which is great for deep understanding - and (for most people) more difficult at the same time. YouTube playlist for this course.
      • Deep Learning Specialisation (5 courses) by deeplearning.ai via Coursera (free audit per course without certificate, or monthly subscription). Python-based, building many tools from scratch.
      • TensorFlow Developer Professional Certificate (4 courses, includes DL, TensorFlow, computer vision, NLP) by deeplearning.ai via Coursera (free audit per course without certificate, or monthly subscription). These two series (TensorFlow in practice, and DL specialization) seems to be a good basis for more advanced applications.
      • IBM Machine Learning Professional Certificate (6 courses, includes DL and reinforcement learning), IBM via Coursera (free audit per course without certificate, or monthly subscription; currently, the 1st month is free via this IBM offer). Intermediate level.
      • Advanced Data Science with IBM Specialization (4 courses, the 4th one is a capstone project, so, you'll have a project to include in your portfolio), IBM via Coursera (free audit per course without certificate, or monthly subscription; currently, the 1st month is free via this IBM offer). Advanced level (they say "Designed for those already in the industry", includes practical use of the models, scalability...). Currently, the 1st month is free via this IBM offer.
      • Professional Certificate in Deep Learning by IBM via edX (6 courses, the last one is a capstone project, so, again, one for your portfolio; incl. DL, NN, using Keras, PyTorch, TensorFlow; applications to computer vision, NLP etc; scaling and acceleration; free audit per course available; whole series with a certificate for 406 euro at this moment)

      NLP:
      Computer Vision:
      • Computer Vision at Kaggle (image classifiers with TensorFlow and Keras)
      • Computer Vision Basics by University at Buffalo / The State University of NY via Coursera (with MATLAB; free audit, certificate = 41 euro)

      Creative/GAN (computer-generated image/video/voice):

      Even deeper to DL/AI/Reinforcement Learning:
      • Reinforcement Learning Specialization (4 courses, the last one is a capstone project) by University of Alberta via Coursera (free audit per course without certificate, or monthly subscription).
      • AI for Medicine Specialization (3 courses) by deeplearning.ai via Coursera (free audit per course without certificate, or monthly subscription). Intermediate level, optimaly preceded by the Deep Learning specialization.

      More resources / various:
      • https://ai.google/education
      • Deep Learning Drizzle: links to about 300 courses in ML/DL/Computer Vision/NLP etc
      • CognitiveClass.ai: free courses, digital badge (some lessons are identical to parts of IBM courses at Coursera)
      • Omdena - "AI for the Real World"-projects: several times per year, Omdena starts projects to solve diverse real-world problems for various organizations; each project takes 8 weeks from the very start to the presentation of the results; several teams work on each of the projects, everybody (even beginners with just a basic knowledge) can apply to become a member of a team, this can be a nice project / item for your CV

      How to audit a course via Coursera for free: Series (specialization and professional certificates) are only available via monthly subscriptions. However, it is possible to audit the most courses (even those included in series). On the website of a course, click on Enroll for Free, but don't choose Start Free Trial! (That's would allow you a 7-day trial period after which you have to pay). Instead, in the lower left corner of the context window, click on Audit the course. This gives you limited access to the course materials (graded assignments are mostly not included, but videos and reading are); it is also temporaly limited but you can audit the same course several times - just don't forget to take notes outside Coursera.

      Let's enjoy learning :)

      ------------------------------
      Lenka Cizkova
      ------------------------------

      Original Message


    • 7.  RE: Best part in Data and AI

      Posted Mon November 16, 2020 01:46 PM
      Hi all, 
      The best part of Data and AI in my opinion is Machine Learning.because it has a much wider coverage and application.



      Original Message