• Type of project: HiWi/Research Project
  • Contact: cemil.emre.ardic@izfp.fraunhofer.de

image

Embrace the future with Fraunhofer Institute.

Explore, Experiment, Excel in Deep Learning!

At Fraunhofer Institute, we offer you the opportunity to dip your toes in the exciting waters of Deep Learning. This isn’t just an industrial setup; it’s a collaborative research institute, a blend of academia and industry, fostering innovation and nurturing talents.

Our platform will allow you to:

  • Dive Deep into Deep Learning: Explore the practical applications of the latest Deep Learning algorithms and be a part of the cutting-edge research that’s shaping our future.
  • Real World Data: Work with genuine, real-world industrial data, and experience the thrill of solving practical, meaningful challenges that can drive our world forward.
  • Bridging the Industry-Academia Gap: Join us and be at the heart of an unprecedented collaboration between industry and academia, fostering an ecosystem that thrives on knowledge, innovation, and mutual growth.

Deep Learning

So what exactly is Deep Learning? It is an exciting frontier in technology, using sophisticated algorithms to create artificial intelligence systems that learn and adapt, just like our brains do. By diving into this captivating realm, you get the chance to teach machines to recognize patterns, interpret language, and even create artistic masterpieces. In deep learning, you don’t just solve problems; you shape an evolving technology that’s rewriting the rules of how we interact with the world. The knowledge you acquire will be at the intersection of mathematics, computer science, and philosophy, opening up limitless possibilities for innovation. By joining us in exploring deep learning, you’ll be embarking on a thrilling journey that puts you at the forefront of the future.

Neural Network Structure

You will be enjoying working with different types of Neural Network Structures. After pre-processing and feature extraction steps, the most important part of deep learning is to design the architecture of the neural network. This will largely depend on the nature of the given tasks, such as classification, regression, generation, etc.

  • Layer Architecture
  • Number of layers and the number of nodes in each layer

  • Overfitting-Performance Trade-off

  • Activation Functions
  • The ReLU (Rectified Linear Unit)

  • Softmax

  • Optimization and Loss Function
  • Stochastic Gradient Descent (SGD)

  • Adaptive Moment Estimation (Adam)

  • Regularization
  • Dropout and early stopping techniques

  • Weight decay (L2 regularization)

  • Selecting Model Type
  • Convolutional Neural Networks (CNNs)

  • Recurrent Neural Networks (RNNs)

  • Hyperparameter Optimization

Requirements

Don’t worry if you don’t know all the things above. These are the fun part ahead. Only fundamental knowledge you should bring with you:

  • Signal Processing
  • Basic Machine Learning/Deep Learning
  • Time-Frequency Analysis

References

  • LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444 (2015). https://doi.org/10.1038/nature14539
  • Ren, Pengzhen, et al. “A comprehensive survey of neural architecture search: Challenges and solutions.” ACM Computing Surveys (CSUR) 54.4 (2021): 1-34.

Interested? Here is my contact details, don’t hesitate to reach me!

  • M.Sc. Cemil Emre Ardic
  • cemil.emre.ardic@izfp.fraunhofer.de