Artificial Intelligence—and machine learning in particular—have come a long way since their early beginnings. The widespread availability and affordability of powerfu
Sequential recommender systems are based on sequential user representations for a given user and sequence length. Each sequence consists of several items in temporal order. Sequential recommender systems aim at exploiting the temporal information that is hidden in the sequence of item interactions of the given user.
Since the invention of the internet, the availability and amount of information has increased steadily. Today we are facing problems of information overload and an ov
In this article I describe the steps and approaches to image recognition for receipt digitalization using computer vision. This is the basic functionality behind apps such as Google Lens, Evernote, PaperScan and taggun.io.
“Would you like the receipt?”—It’s hard to say no to that. Not because you actually want it (you may even throw it in the trash before exiting the store), but because
We extend state-of-the-art sequence-to-sequence neural networks for summarization of long text across windows. By learning transitions, we are able to process arbitrarily long texts during inference.
This blog post describes my master thesis "Abstractive Summarization for Long Texts". We’ve extended existing state-of-the-art sequence-to-sequence (Seq2Seq) neural net
The idea behind the model-agnostic technique LIME is to approximate a complex model locally by an interpretable model and to use that simple model to explain a prediction of a particular instance of interest.
This is the second part of our series about Machine Learning interpretability. We want to describe LIME (Local Interpretable Model-Agnostic Explanations), a popular t
This blog explains the basic concept of Reinforcement Learning, giving you an understanding of the closed loop system, in which an agent uses actions to change the state of the environment and thus receives rewards, with the goal of maximizing the return.
I would like to start this series about reinforcement learning by giving an overview of what reinforcement learning is, what it is used for and what terminology is ne
Entropy is a significant, widely used and above all successful measure for quantifying eg. inhomogeneity, uncertainty or unpredictability. It is an integral part of the latest machine learning models deployed on real-world data sets. In this article, I want to highlight the simplicity, beauty and meaning of entropy.
If you are dealing with Statistics, Data Science, Machine Learning, Artificial Intelligence or even general Computer Science, Mathematics, Engineering or Physics, you
Despite their outstanding performance on various tasks, machine perception systems are not infallible. We highlight this problem by means of particular adversarial glasses that manage to force face recognition systems to make mistakes und we show how to achieve robustness against such attacks.
Despite the fact that machine perception systems achieve superhuman performance on different perceptual tasks, researchers have recently demonstrated that they are no
Everybody talks about AI and deep learning and everybody uses it, including you! But what exactly is deep learning and what are artificial neural networks? In this article I shine a light on some basic yet crucial concepts in an attempt to lift the veil.
Artificial intelligence or deep learning: Everybody talks about it and everybody uses it, including you! Of course you immediately have the evil terminator in mind wh
I use state-of-the-art NLP techniques to improve an existing pricing model in an online car market. Online car markets usually use technical car attributes for price prediction with sellers adding description texts to provide more details. In my thesis, I use these texts to improve the existing pricing model.
tl;dr: This blog post summarizes my masters‘ thesis. I use state-of-the-art NLP techniques to improve an existing pricing model in an online car market. Online
This article presents SeqPolicyNet, our Deep Learning approach to accessing information stored in an Elasticsearch instance given natural language questions.
tl;dr (spoiler alert): We’ve trained an advanced neural network to query Elasticsearch based on natural language questions. Our model, called SeqPolicyNet, incorporat
This article unveils the connections between artificial intelligence, machine learning and deep learning based on a simple example. It suits as an introduction for newbies as well as a reference point for advanced readers looking for more complex content.
There has always been a gap between the capabilities of men and machine. While computers were able to perform complex multiplications or store large amounts of data,
Machine learning as a service (MLaaS) bietet Unternehmen eine einfache Möglichkeit, Daten zu verarbeiten, Modelle zu trainieren und Prognosen zu erstellen. In diesem Artikel werden die Angebote von vier der größten Cloud-Anbieter vorgestellt: GCP, AWS, MS Azure und IBM Cloud/Watson.
Cloud Computing gewinnt durch sein flexibles Bereitstellungsmodell immer größere Bedeutung. Von Software (SaaS),Plattformen (PaaS) bis hin zur IT-Infrastruktur (IaaS)
In the last years, artificial neural networks (ANN) have successfully been applied across a number of tasks. However, designing well performing ANNs requires expert knowledge and experience. Neuroevolution aims at solving this difficult and often time-consuming process by using evolutionary techniques.
In the last years, artificial neural networks (ANN) have successfully been applied across a number of tasks, such as image classification, speech recognition and natu
In this article we will look at the history of Neuroevolution and present state-of-the-art work that was performed by Google, Uber and other companies.
Neuroevolution describes the evolution of Artificial Neural Networks for problems in the domain of supervised or reinforcement learning. This article is the result of