I’m happy to introduce you to our newest data science team member: machine learning and pattern recognition aficionado Vincent Spruyt. As Head of Deep Learning, he will be heading up the application of deep learning methods for the Argus data model and intelligence.
Vincent pursued his PhD research in Computer Vision, and is an expert on the development of self-learning and context-sensitive algorithms. This knowledge he gladly shares on his VisionDummy blog. Vincent graduated as a Master of Science in electronic engineering, and obtained bachelor degrees in both applied informatics and electronics-ICT. His past as a Java JEE architect at Toyota Motors Europe allowed him to get acquainted with business intelligence and data mining, and encouraged him to focus on deep learning architectures and convolutional neural nets.
Hi Vincent, we’re thrilled to have you joining the Argus Squad. I’m assuming you’re just as excited?
VS: Like many of us, I’ve always believed that humanity would one day be served by a true Hal 9000 (Space Odyssey). Joining a team of like-minded enthousiasts that would like nothing more than to make this happen, is indeed exciting at the very least.
You’ll be part of the machine learning team. As such, could you explain to me what’s the difference between machine and deep learning?
VS: Machine learning is a field of Artificial Intelligence that deals with pattern recognition and data interpretation. Since machine learning’s rise in the late fifties, the techniques have significantly advanced, yet they have not been able completely fulfil the high expectations. More recently, in 2006, researchers have found a way to efficiently train deep neural architectures, inspired by the work of cognitive neuroscientists. These architectures closely mimic our understanding of the brain’s neocortex. They allow us to develop algorithms that are able to autonomously learn extremely complicated non-linear data dependencies, without human intervention. In fact, deep learning architectures have been shown to outperform any existing learning method, and could be considered a revolution in the field of artificial intelligence.
Ten years from now, what magic will you be working on? What seems like the most incredibly fascinating thing to make or do, that is not quite possible – yet?
VS: I suppose this is the question were I should have introduced Hal 9000. True artificial intelligence has been blocked in the past, by the lack of data, the lack of efficient learning algorithms, and the lack of computational power. Today the impossible seems to happen; big data becomes ubiquitous, deep architectures allow for unsupervised learning, and cloud computing provides us with a virtually infinite amount of processing power. Ten years from now, we will be educating the artificial intelligence that is being built today, using its beauty to make the world a better place.
Your tutorials on the Vision Dummy blog about computer vision and machine learning get some very enthousiast responses. Can we expect more of these in the future?
VS: Of course! Although extremely fascinating, the field of machine learning can sometimes be difficult to grasp due to its mathematical foundations. However, complicated mathematical formulae are often more of interest to academic researchers than to programmers and entrepreneurs. With my blog I try to unleash the power of machine learning to the community of programmers that have built and supported the internet we know today. Imagine what they can do, over the next ten years…
If we allow you access to our joint music playlist, what type of music will it be playing more often?
VS: Haha, does this question mean that an incorrect answer will not grant me access to the joint playlist? I’m into folk, indie rock, and the singer-songwriter genre. I love Leonard Cohen, Okkervil River, the Smiths and Neil Young.
Anything we should know I forgot to ask you about?
VS: We should know the ultimate question of life, the universe, and everything (The Hitchhiker’s Guide to the Galaxy). Trying to find the question when the answer is given, is what machine learning guys call ‘inference’.
You’re the second Vincent joining the team. So, if it were up to you, what would your preferred nickname be?
VS: Honestly I’m the worst when it comes to remembering names. While it’s probably not the most polite solution, I often try to avoid calling people by name. Maybe you can just do the same 😉 Or we could use numbers instead of names, in which case 42 is a pretty cool number.
Vincent ’42’ Spruyt commences working with our data and mood detection systems this Monday, June 1st. After deliberations, we’ve also decided to grant him access to the joint Argus playlist. If you want to talk to 42 about data science, deep learning and neural architectures, or sci-fi movies, you can reach him on Twitter (@vincent_spruyt), by email (firstname.lastname@example.org), or head over to his blog at visiondummy.com.