This article is an effort to make you into a “semi-expert” in artificial intelligence, cognitive computing, deep learning and neural networks from scratch. Here I will share a few cool learning resources for these topics. These resources include documentaries, TED talks, online lecture videos, and books. There are several videos and online books included in this post to help you learn these concepts. These resources vary from introductory to advanced learning. These learning resources are organized under the following four categories
1) Cognitive Computing and Artificial Intelligence – you will find some cool documentaries and TED talks in this section to get introduced to the fascinating world of AI and cognitive computing.
2) Neural Networks Basics – neural networks are at the core of deep learning, cognitive computing, and AI. In this section, I will share a few videos to demystify and learn neural networks.
3) Deep Learning Fundamentals – In this section you will learn the fundamentals of deep learning from the best in academia i.e. researchers from Stanford, Oxford, Montreal, and Toronto.
4) Neural Networks and Deep Learning Practice Sessions – finally time for you to practice these new skills through working examples. Make sure you brush up your Python before this using this link – Book to learn Python.
However, before we jump towards these resources let’s visit the wonderland of data science which these fields are part of.
Wonderland of Data Science
The Wonderland of Data Science – by Roopam
The period between 1905-1930 was the golden age of physics. Einstein, Bohr, Schrodinger, and other great minds were working relentlessly on some of the greatest ideas in physics including the General Theory of Relativity, and Quantum Mechanics. These ideas shaped the direction for physics for the years to come. Similarly, Watson and Crick had ignited the golden age for biology and genetics in 1953 by discovering the double-helix structure of the DNA molecule. For chemistry, a similar golden age was in the late 19th century initiated by Dmitri Mendeleev’s periodic table.
Every field of science goes through a golden age where a myriad of novel ideas is generated. These ideas not only shape the future of the field but provide a quantum jump in human understanding about their surroundings and nature. We are undoubtedly living in the golden age of data science, statistics, and machine learning. As a practitioner of data science for all these years, I will be honest, I feel like how Alice must have felt in Wonderland. There are new and fascinating ideas coming from everywhere. This statement by Alice explains the reality of being in Wonderland.
I knew who I was this morning, but I’ve changed a few times since then.
–Alice (Alice in Wonderland)
Evolution of Data Science
Risk and marketing models were the pioneers of data science in industry. Like most professionals of data science, I started my career building these models and still do. Through them, I have learned a great deal about sociology and social-psychology. These models are still extremely prominent for most industrial applications. But one of the key features about Wonderland is that it is always changing. In the wonderland of data science, one of the new directions is cognitive computing and artificial intelligence. AI has been around in movies for a long time. However, it is becoming a reality now with the advent of machine learning.
There is a faction that will argue that machine learning & cognitive computing is not data science. The reason for their belief is since some machine learning models are like black boxes. These black box models don’t explain intuitive cause and effect relationship between predictor and target variables. I think this is a highly myopic view. This is like saying Quantum Mechanics is not physics or the General Theory of Relativity is not physics since they have different approaches to understand physical phenomena. To me, data science is an act of generation of knowledge/intelligence through data/experience. Hence, both causal models and black box models are part of this wonderland of data science.
In this article as promised earlier, I am going to share some cool learn resources for artificial Intelligence, cognitive computing, and deep learning. Neural networks are at the root of these topics. Please also read an earlier article on YOU CANalytics about artificial neural networks. Neural networks are inspired by our brain and are referred as black box models. Let me explain how black box models work.
Chicken Sexing & Neural Networks – Black Box Models
Imagine you run a large commercial poultry farm. For a purely commercial purpose, it is absolutely important for you to distinguish between male and female chick right after hatching. You will keep all the female chicks for the production of eggs and a few male chicks. Most of the other male chicks will be killed immediately afterward. Now, the task of distinguishing male and female chicks is not at all easy because they all look almost exactly the same. It requires a trained professional called ‘chick sexer’ – I am not kidding that’s a real name for the profession. Chick-sexers use a method called vent sexing. This requires a close examination of the chicks’ rear to classify them as male or female.
A master chick sexer is extremely accurate in classifying sex. However, even the best chick sexers won’t be able to explain their method for you to learn. Chicken sexing is a black box model where chick sexers do their job perfectly without knowing how they are doing it. The training for an apprentice chick sexer is equally fascinating. The master chick sexer overlooks the apprentice while the apprentice performs his task of distinguishing chicks. The master just provides a feedback of just yes or no for each distinction the apprentice makes. This simple feedback loop and a few months of rigorous effort, hardcodes the art of chicken sexing in the apprentice’s brain. Now the apprentice is ready to take over from his master without knowing how he did it.
Knowledge and Awareness
This creates a clear distinction between knowledge and awareness. It is not necessary that all knowledge leads to awareness. Similarly, some data science models (cause and effect models) produce awareness or provide reasons for how things work. The other data science models, like neural networks, produce knowledge without awareness. They are black box models like chicken sexing. Neural networks are at the core of deep learning algorithms. Moreover, deep learning algorithms are at the core of most of the modern cognitive computing and systems with artificial intelligence.
In the following sections, I will introduce learning resources for artificial intelligence, cognitive computing, deep learning, and neural networks.
Artificial Intelligence and Cognitive Computing – Get Excited
Cognitive computing is a subset of a broader field called artificial intelligence. Cognitive computing facilitates applications that are classified under artificial intelligence. At present, the toughest problems in cognitive computing involve tasks that humans can perform effortlessly such as understanding languages, images, sound, and emotions.
AlphaGo Zero: Starting from scratch
– DeepMind’s Professor David Silver describes AlphaGo Zero
YOU CANalytics Rating (4 / 5)
Run Time: ~2:13 mins
Unlike real-life, games have well-defined rules. This makes them a great way to improve machine learning algorithms and cognitive computing. It all started with Deep Blue when it defeated Gary Kasparov in a chess duel. Then came IBM Watson which defeated the human champions of Jeopardy. AlphaGo Zero is the latest and most powerful addition to this list. Unlike Deep Blue and Watson, AlphaGo Zero learns on its own by simulating games and data. This is explained in this 2 minutes video by Professor David Silver of DeepMind. If you are interested in learning details about AlphaGo Zero algorithms then you could find it in this paper.
Smartest Machine on Earth – PBS Nova Documentary
YOU CANalytics Rating (5 / 5)
Run Time: ~70 mins
Jeopardy! is a television game show. Brad Rutter and Ken Jennings are among the best Jeopardy! winners of all time. IBM Watson won the Jeopardy! the challenge against these former winners. It is a big deal for a machine to beat humans in such a complicated game. Watson initiated the age of cognitive computing. Do watch this PBS-Nova documentary to understand what Watson does and how. PBS-Nova has a history of producing top quality science documentaries and this one is no exception.
Future Intelligence – Next World Discovery Documentary
YOU CANalytics Rating (5 / 5)
Runtime: ~44 mins
This documentary ventures into the future where artificial intelligence and intelligent computer systems will be a norm. It is a great sneak peek into the future of machine intelligence and what the future holds.
How We Teach Computers to Understand Pictures– TED Talk
YOU CANalytics Rating (5 / 5)
Runtime: 18 mins
Fei Fei Li’s 3-year-old son Leo is the star of this wonderful TED talk by her. She loves her research and it is always fun to listen to people who love their job. You will learn a great deal about image processing and image analytics in this TED talk. By the way, Leo, a normal 3-year-old kid, can beat the best computer systems when it comes understanding pictures. Ms. Li will explain what it will take for computers to understand images the way humans do.
Cognitive Computing – TED Talk
YOU CANalytics Rating (3.5 / 5)
Runtime: 17 mins
This is a good talk to understand the current state of cognitive computing. If the above talk is inspirational this one is informative.
Neural Networks – Start with the Basics
How to make Neural Networks in your bedroom– TED Talk
YOU CANalytics Rating (4.5 / 5)
Runtime: 8 min
Brittany Wenger is the winner of the Google Science Fair in 2012. She had built a neural network based system that can detect malignant tumor for breast cancer at 99% accuracy. This is her story.
Neural Networks Demystified – Welch Labs
YOU CANalytics Rating (4.8 / 5)
Runtime: ~27 mins (Total 7 Part series)
The following lecture series will get you started with the basics of neural networks. You can find all the 7 parts of this series in the index at the top left corner of the video. This is among the best technical yet simple introductions to neural networks.
Mathematics of Neural Networks – NPTEL
YOU CANalytics Rating (3.5 / 5)
Runtime: 60 mins
It’s nostalgia for sure. This is how I was taught advanced mathematics and probability in IIT Bombay. Possibly, not the best way to learn neural networks but 8 years ago this video helped me understand the mathematics of neural networks. One of the comments on YOU TUBE calls this professor an owl in human disguise – I will leave it up to you to make your own wise judgment.
Deep Learning – Let’s get Technical
University of Toronto, Stanford University, Université de Montréal and Oxford University are the best learning centers and research powerhouses for deep learning. In this section, you will learn from the best when it comes to deep learning.
Neural Networks for Machine Learning – Jeffery Hinton (Coursera / Univesity of Toronto)
YOU CANalytics Rating (5 / 5)
Runtime: ~ 750 mins
It’s time to learn deep learning from the godfather of the field Jeffery Hinton. He continued to work on neural networks when the world had given up on them. He is the pioneer of deep learning and thanks to Coursera for bringing this wonderful lecture series to us. Find the entire series in the index at the top left corner of this video.
Andrew Ng on Deep Learning– Lecture Video : Stanford University
YOU CANalytics Rating (5 / 5)
Runtime: 46 min
Andrew Ng is another giant in the field of deep learning and machine learning. You should check out his lectures on machine learning on Coursera if you already haven’t. In this video, he talks about deep learning and sparse coding.
Lecture Series on Deep Learning – University of Oxford
YOU CANalytics Rating (4 / 5)
Runtime: ~ 1000 mins
This is yet another lecture series on deep learning. This time from Oxford University. Make sure you catch-up the 10th and 12th lecture on convolutional and recurrent neural networks.
Deep Learning Summer School 2015, Montreal: this link contains several talks on deep learning from some of the leading minds in the field. These talks were part of summer school at Montreal in 2015.
Neural Networks and Deep Learning – Practice Sessions
Now that you have understood the fundamentals of neural networks and deep learning you are ready to get some hand on exposure to these fields. Brush-up your Python before you start with these tutorials and online books.
This is a great online book, available for free, to learn and practice fundamentals of neural networks and deep learning. This book has 6 chapters. The book will introduce you to some fundamental concepts in these topics which are followed by working examples in Python.
Deeplearning.net is another fantastic site to practice your learning of deep learning. The site has several tutorials for you to brush up your skills and become a quasi-expert in deep learning.
On this site, you will find a quick tutorial of convolutional neural networks for human facial keypoint detection.
Sign-off Note
Hope you learn deeply to gain a deeper understanding of deep learning. See you soon with a new article.
5 thoughts on “Learning Resources : Artificial Intelligence, Cognitive Computing, Deep Learning, & Neural Networks”
Hi Roopam,
Can you please also suggest some good books which give mathematical as well as intuitive interpretation of new age ML algorithms such as ensembles, SVMs etc? I feel like we build these models (R libraries help a lot) but there are still many gaps in real understanding.
Hi Roopam,
Can you please also suggest some good books which give mathematical as well as intuitive interpretation of new age ML algorithms such as ensembles, SVMs etc? I feel like we build these models (R libraries help a lot) but there are still many gaps in real understanding.
Thanks,
Rishabh Soni
Try “Pattern Recognition and Machine Learning” by Christopher Bishop
Excellent source of exciting information. Thanks for compilation.
Excellent information on exciting topics.
Good page and resources for getting into the AI field. Great initiative, will follow
your content. Good luck!
/Robert Dreamrealist
Community Manager @
https://www.facebook.com/SwedishAISociety/
https://www.facebook.com/AIAware/