Can Neural Networks Learn From Their Nightmares?

In your dreams, your mind processes your experiences in strange ways. Artificial neural networks are often compared to human brains. Is there a way to make neural networks “dream nightmares”?

Required interest(s)

  • Neural networks
  • Image recognition
  • Artificial Intelligence
  • Machine Learning

What do you get

  • A challenging assignment within a practical environment
  • € 1000 compensation, € 500 + lease car or € 600 + living space
  • Professional guidance
  • Courses aimed at your graduation period
  • Support from our academic Research center at your disposal
  • Two vacation days per month

What you will do

  • 65% Research
  • 10% Analyze, design, realize
  • 25% Documentation

Dreams are typically related to something you have experienced during the day. But in your dreams, those things are changed in strange ways. For instance, if you have seen one small spider sitting in a corner, you may dream about really big spiders running after you. As weird as they may seem, dreams may help your brain process and learn from your day.

Artificial neural networks are often compared to human brains. Following this analogy, AI researchers have found ways to extract dream-like images from neural networks. These “dreams” look like strange modifications of images that the neural networks processed during their training phase. For example, take a look at: Google AI Blog: Inceptionism: Going Deeper into Neural Networks (

In another strain of research, researchers have shown that neural networks can be fooled into making wrong decisions by modifying their input by a very small amount. For instance, when a researcher cleverly changes only a few pixels in an image of a panda, the network thinks it is an image of a monkey. This is called an adversarial attack. For example, take a look at: When AI Becomes an Attack Surface: Adversarial Attacks | Computer Science Blog (

Can we combine these two research areas? For instance, can neural networks “dream nightmares”? (if their dreams are adversarial attacks)? Or, can we perturb their dreams such that they become nightmares (adversarial attacks)? And, most importantly, can neural networks learn from these nightmares, so that we can improve them and make them more resistant against uncertainty or attacks?

Your task would be to do exciting research in the area of neural network dreams and adversarial attacks. After reading up on the relevant literature, your first step would be to find a way to make neural networks “dream nightmares”. The next objective is to see if you can improve the neural network using these nightmares, namely to answer the question: can the network learn from its nightmares? Of course, the goal would be to significantly improve the network, which means the quality of the network must be measured.

You have some freedom to fill in this research according to your own interests.

About Info Support Research Center

We anticipate on upcoming and future challenges and ensures our engineers develop cutting-edge solutions based on the latest scientific insights. Our research community proactively tackles emerging technologies. We do this in cooperation with renowned scientists, making sure that research teams are positioned and embedded throughout our organisation and our community, so that their insights are directly applied to our business. We truly believe in sharing knowledge, so we want to do this without any restrictions.

Sign up for this assignment

  • Toegestane bestandstypen: docx, doc, txt, pdf, Max. bestandsgrootte: 8 MB.
  • Toegestane bestandstypen: docx, doc, txt, pdf, Max. bestandsgrootte: 8 MB.

Application procedure

  1. 1
  2. Introductory meeting

    Discuss (study) career, interests and ambitions and introduction Info Support.

  1. 2
  2. Review

    Assessment of professional knowledge and personality (capacity, competences and motives).

  1. 3
  2. Selection interview

    Deepen professional knowledge and personality.

  1. 4
  2. The signing of a contract

    Contract offer and invitation for drawing moments.