"Machine Learning" sounds like miracles and wonders, like robots behaving like cute little kids maybe ... but it is not (yet) like this. ML is a collection of very powerful algorithms and algorithm design methods which is firmly rooted in **probability theory, statistics, linear algebra and some calculus**. ML is one of the most mathy areas of Computer Science. The input to an ML algorithm is (large quantities of) data - for example images, video, geo-surveillance data, scientific measurements, financial records, web-crawling harvests. The ML algorithm** tries to find regularities (structure, redundancies, symmetries...) in the data **and use these to create a formal representation of the data (a **model**) in a compressed format which should be much more compact than the original data. We may be talking about real large items here: current state-of-the-art ML algorithms may be fed with many billions of data points and "compact" them into a representation with millions of parameters. After having **learnt** a model from the data, the model can be exploited for a variety of purposes, for instance data **interpolation or prediction, pattern classification, control, or pattern generation**.

If you are contemplating choosing ML for your Guided Research thesis, the first thing for you is to find out whether you really like this kind of stuff and whether it is accessible to you. The standard way to find out is to take the 2nd year IMS course on machine learning and emerge from that course with some enthusiasm. If you haven't taken that course, you will face a rather steep working-in challenge because you will have to self-study the basics of machine learning (recommendation: my online lecture notes). At any rate, you should feel good friends with maths, especially linear algebra. -- Students of mathematics usually have little difficulties absorbing the requisite background knowledge and are very welcome to engage in a machine learning project!

The next thing is to get an idea of the specific research that is carried out in the MINDS group at Jacobs. Your thesis work will be closely connected to it. There are three main lines of research:

- Investigations of recurrent neural networks of the "Echo State Network (ESN)" type. To get a very first quick impression, take a look at the echo state network page on this MINDS website. Computer Science students typically embark on this line of work.
- Investigations of mathematical models of stochastic symbol sequence data (like texts or DNA sequences), using "Observable Operator Models" (take a first look here). This is a math-intense area of study and typically students of ACM or mathematics go for it (and like it!).
- Recently a new topic has been added, conceptors. This builds on echo state networks in some ways but can also be seen independent of ESNs. Conceptors are "neural filters" which can be learnt from experience and which can make a neural network generate complex patterns in a controlled way, for instance human motion patterns. Since conceptors are an extension of Echo State Networks and rather more mathematically / conceptually advanced, this topic would likely be suitable only for particularly ambitious students who are prepared to spend a significant extra effort (reward: a particularly advanced state-of-the-art thesis).

To get a clearer picture of the first kind of research (echo state networks), please read the following papers in the given order:

- A Scholarpedia article for a first overview.
- A short highlight paper: H. Jaeger and H. Haas,
*Harnessing Nonlinearity: Predicting Chaotic Systems and Saving Energy in Wireless Communication.*Science 304, 2 April 2004, pp. 78-80 (preprint pdf) - An easy overview paper of the status of echo state network research in the current landscape of ML: M. Lukoševičius, H. Jaeger, B. Schrauwen (2012):
*Reservoir Computing Trends***.**KI - Künstliche Intelligenz, 1-7 (Preprint pdf) - An introductory, more detailed technical report which has a large number of examples in it: H. Jaeger (2001):

You might also wish to check out a (very good) BSc thesis written on an ESN theme (by Valentin Vasiliu, 2016).

If you feel you might become interested in the observable operator model work, read the first part in the 2-part paper

- H. Jaeger, M. Zhao, K. Kretzschmar, T. Oberstein, D. Popovici, A. Kolling (2006): Learning observable operator models via the ES algorithm. In: S. Haykin, J. Principe, T. Sejnowski, J. McWhirter (eds.), New Directions in Statistical Signal Processing: from Systems to Brain. MIT Press, Cambridge, MA., 417-464 (draft version, pdf)

Note: OOM theses written in the MINDS group seem to be good for carving an academic carreer path: Cristian Danescu-Mizil (Master thesis 2007) is now an assistant professor at Cornell, Anca Dragan (BSc thesis 2009) is assistant professor at Berkeley, and Josip Djolonga (BSc thesis 2011) is a PhD student at ETH Zurich with a Google European Doctoral Fellowship. You can check out Josip's BSc thesis if you want to get an OOM feeling.

Finally, research on conceptors is young and there is not much literature yet. A short "teaser paper" (11 pages) is

- H. Jaeger (2014):
*Conceptors: an easy introduction*. (arXiv)

So far there has only been a single conceptor-based BSc thesis, - a fine one -, by Alina Dima (2014).

If you are done with this reading (of either of the three directions), we can start talking business... I don't have ready-made thesis topics. Rather, I'll be sitting together with you and define a thesis topic which is tailored to your interests and background.