One of the best way to get better at machine learning and deep learning is to watch a lecture from an expert and work your way along with it. If you do so, you get the best of both the worlds – you learn from the experts across the globe and also get hands on knowledge.
In this article, I have provided a list of YouTube videos, which you can use to improve your knowledge in these areas.
How should you watch them ?
You’ve got to follow a ritual (Just Kidding!). For your ease, I have created a ‘to be followed’ sequence / order of these videos. I have categorized the videos into: Machine Learning, Neural Networks and Deep Learning. If you are new to this, I’d recommend you to follow the sequence for better understanding. If not, feel free to follow your own route and let me know, how it turns out.
If you feel that the list is overwhelming, take one byte at a time! Consider this as a collection of references to be consumed over time. These videos range from a few minutes to hour long videos. For your convenience, I have also mentioned the summary against each video for the overview purpose. Go ahead and check them out.
Videos related to Machine Learning
Summary: What better way to start this journey than to hear from one of the best machine learning teacher & expert across the globe. In this video, Andrew Ng talks about his childhood dream to build robots that can actually think and work like humans and improve the lives of millions. He talks about the similarity in human brain and the software in machines that can make them function like humans.
Summary: This is the complete playlist for the lectures on Stanford Machine Learning (CS229) by Prof. Andrew Ng at Stanford. I personally think this is better than the coursera classes and enjoyed this thoroughly.
In these lectures he covers machine learning concepts which include Linear / Logistic regression, supervised and unsupervised learning, learning theory, reinforcement learning and adaptive control. He discusses techniques like Naive Bayes, Neural Networks, SVM, Bayesian statistics, Regularization, Clustering, PCA and ICA. He also discusses some recent applications of Machine learning such as robotic control, data mining, autonomous navigation, bioinformatics, speech recognition, and text and web data processing.
If you are a complete newbie and want a lucid introduction, which challenges you at times, go with these videos
Summary: In this Caltech’s Machine Learning Course – CS 156, Professor Yaser Abu-Mostafa gives in-depth details of various machine learning concepts and techniques. This course is very heavy on mathematics and the theory behind Machine Learning with a bit difficult programming exercises. It balances theory and practice, and covers the mathematical as well as the heuristic aspects. The lectures follow each other in a story-like fashion. You should do this course for its challenging assignments. This lecture series has 18 videos.
Summary: In this awesome video, Tavis Rudd talks about his two-year long journey which led him to create this cool feature of coding using Voice Recognition. He did a lot of vocab tweaking and duct-tape coding in Python and Emacs Lisp to develop this system which enables him to code a lot faster. He gives a live demo of the software that he has developed and makes his system do things in seconds through his voice which otherwise may take hours to code!
Summary: In this video, Mr. Stephen Hoover talks about a cloud based data science platform for data analysis built using Python at his company Civis Analytics which helps analysts to perform their work much faster and with much little effort. He talks about various machine learning aspects of the platform and talks about some open source libraries in Python which help in data analysis like Pandas, NumPy and Scikit-Learn.
If you have survived the journey till now, Congratulations! Have a look at the next exciting video and you will be all prepped up for the next 2 sections, i.e. Neural network and deep learning! The fun doubles up, if you were a Mario fan as a kid.
Summary: This video shows how a computer program called MarI/O learns to play Super Mario game. This program is made of Neural Networks and Genetic Algorithms. The video showcases the actual biological evolution of the program compared to the human brain. This is one of the awesome applications of Machine Learning which demonstrates the scope of Machine Learning in various activities that humans can perform.
Videos related to Neural Networks
Summary: Here’s a playlist known by Neural Network Class. It covers the basic to advanced level concepts of neural network ranging from artificial neuron, activation function to recursive network training. You won’t find long duration videos in this playlist. The videos are short and crisp with the longest duration of 24mins. I’d recommend this to every beginner learning neural networks.
Duration- 12:40 mins
Summary: This series of video teaches how to train a neural network i.e. neural network training. The way of explaining is really good. This particular video gives the overview of complete training process. Once you click on this video, in the ‘Up Next’ section on YouTube, you’ll find its subsequent videos covering topics such as neural network error calculation, gradient calculation etc which are indeed helpful.
Summary: ANN exploits the non linearity. It assists the process of input-output mapping says Professor S. Sengupta, IIT Kharagpur, India. He flawlessly explains the concept of artificial neural network in the simplest possible manner. He explains using a pen and paper which actually helped me to understand these concepts. Towards the end of this video, he touches upon the applications of ANN. Don’t miss out the ‘Up Next’ section.
Summary: Convolutional neural networks are used to recognize objects, images and videos. In this 47 minute video, you’ll be introduced with the concept of deconvolutional network followed, the insights for architecture selection in the convolutional networks. The role of visualization is to present an insights on the performance of each layer using which improvements can be made.
Summary: A person who needs no introduction, Geoffery Hinton, gives an enriching talk on neural network at GoogleTechTalks. This video will help you to build a solid foundation in deep machine learning. It will also take you through the journey of neural networks from the past to future. Geoff cover topics such as back propagation, digit recognition, Restricted Boltzmann Machine and related topics.
Summary: This short video explains artificial intelligence using evolving neural networks by using a designed program ‘Neural Bots’. The complete activity of the bots have been visualized using a pre-defined set of commands. This will be fun to watch.
Duration- 09:04 mins
Summary: Uploaded by Microsoft Research, this video is a small talk given by Chief Research Officer, Rick Rashid. Rick demonstrates a speech recognition breakthrough using deep neural networks & machine translation which converts spoken english to chinese languages, simultaneously, reducing the amount of recurring errors at every layer.
Duration- 1:04:03 hrs
Summary: This videos covers the unconventional topic of learning from bacteria about information processing. The speaker begins with the part of rudimentary intelligence which involves cognition, sensing, processing. He also shows the pattern of rethinking bacteria. Finally, the use of social networks can be seen driven by chemical tweeting.
Summary: The title of the video is clear enough to explain its content. This video demonstrates the complete process of a gene which learn to jump over the ball.
Summary: Just like previous video, it also emphasizes upon the applications and wide range of implementation of neural networks. In this video, a genetic algorithm learn how to fight. Such videos triggered my appetite to learn as I realized the unscalable potential of neural networks.
Videos related to Deep Learning
Duration- 52:40 mins
Summary: This video teaches the implementation of Deep Learning in Python. It begins with a ‘motivating’ problem of handwritten digit recognition. It also demonstrates the complete codes of python used to solve a dataset based on 60,000 images. The speaker then emphasizes upon the codes and makes sure that he doesn’t miss out on explaining the important set of codes and algorithms.
Duration- 1:09:04 mins
Summary: This video gives a good start to understand the concepts of Deep Learning. Markus begins this talk by explaining the story behind deep learning. Then he gives a quick refresher on linear algebra which is followed by basic neural networks, unsupervised models and RNN-GSN. Later, the explains how to implement simple neural networks in Python using Theano.
Duration- 1:24:06 mins
Summary: This talk will introduce you to a new concept which integrates deep learning and big data. Deep Learning has began to derive significant value from big data. At the later half of this video, you’ll find a useful discussion happening among the research scientists from Google, Facebook and other big giants. Their discussion covers most of the elements of deep learning and big data which are essential to drive its future growth.
Duration- 2:00:04 mins
Summary: This video got published less than a week back. This is the first tutorial I found on computer vision. This tutorials explains the concepts such as (spatial pooling), normalization, image net classification etc. Towards the end, various amazing applications have been displayed using a collection of useful images.
Duration- 50:30 mins
Summary: The Computer Science Department of University of Oxford released this tutorial few months back. This is by far one of the most sought video on convnets. The speaker discusses the concepts of using convnets for object recognition and language, how to design convolutional layers, how to design pooling layers. In the later half of the video, he discusses the process to build convnets in Torch.
Duration- 48:20 mins
Summary: This talk is given by Andrew Ng, Founder of Coursera to address the development of unsupervised feature learning and deep learning that can automatically feature representations from featured data. In this talk, Andrew describes the useful concepts behind unsupervised feature learning and deep learning, describes few algorithms and presents a pertaining case study.
Duration- 1:05:20 mins
Summary: Geoff Hinton, a pioneer in Machine Learning talks about the recent developments in Deep Learning in this video. Laying emphasis on the mathematical aspects of various algorithms, he talks about tasks such as object recognition, information retrieval, and modeling motion capture data in which Deep networks have been quite successful.
Summary: This is an audio version of interview with Geoffrey Hinton. In this interview, he describes how does google implements artificial intelligence system. Also, he emphasizes on the learning component of humans, and machines using neural nets. This is a must watch (listen rather) for every machine learning enthusiast.
Duration- 54:31 mins
Summary: Yann LeCun from the Computer Science Department of the NYU talks about things where Learning Theory is difficult to apply to and presents it as a challenge for the community to study it. He talks about various Deep Learning concepts and his interest in Learning Representations in particular which he thinks is probably the next step for AI Machine Learning.
Duration- 1:05:30 hrs
Summary: This video from the Deep Learning expert at NVIDIA, Mike Houston, talks about the Deep Learning training system called NVIDIA DIGITS, along with the NVIDIA DRIVE PX car computerwork in enabling cars to drive themselves. He talks about the training tools and platforms that his team uses for building these self-driven cars along with the Deep Learning algorithms which they use for the same.
Duration- 56:02 mins
Summary: In this video, Sergey Levine is a postdoctoral researcher working with Professor Pieter Abbeel at UC Berkeley talks about the applications of Deep learning in Decision Making and Control. He focusses on applications like Continuous Control Tasks and some other broader applications at the end. He also describes the algorithms that tackle these challenges using Supervised Learning.
Duration- 1:00:23 mins
Summary: Jeff Dean, senior fellow, Google Knowledge Group talks about how we can build more intelligent computer systems using Neural Networks and Deep Learning. He focusses on the abilities of computer systems such as basic speech and vision, language understanding and user behavior prediction and how has he applied all these techniques at Google in its various products.
Hope you find this list useful. As mentioned before, it may look overwhelming for starters – take one step at a time and code along side. If you have suggestions for videos, which should be added to this list, please feel free to add them here. I will love to watch them!