20 feb, 18:00–19:00

Probabilistic Machine Learning: What it is and How it Enhances Robotic Perception

Are you interested in probabilistic machine learning and wonder how to deal with uncertainty in machine-learning systems? During this lecture at Goto 10, Gabriela Zarzar Gandler, MSc in Machine Learning (KTH), will introduce Gaussian processes to enhance robotic perception.

While data is one of the key elements of machine-learning systems, it is useless unless some knowledge is extracted from it. This can be done through learning, which can be thought of as inferring plausible models that explain the observed data. Typically, a machine uses such models to make predictions about future (or unobserved) data. However, any sensible model will eventually be uncertain when predicting data that hasn’t been observed. So how to deal with uncertainty in machine-learning systems?

In the fascinating field of probabilistic machine learning, uncertainty is modeled by applying tools from probability theory. We will particularly talk about Gaussian processes, a flexible non-parametric data-driven approach for modeling unknown functions. They enable predictions that not only provide estimates for unobserved data, but also information regarding how certain the estimates are. Uncertainty information is a crucial piece of information in many real-life systems, such as in robotic tasks.

This lecture will be divided into two parts. The first part will cover the theoretical concepts and expects a basic knowledge of probability theory from the audience. We will introduce Gaussian processes and discuss examples with one-dimensional as well as three-dimensional data. In the second part of the lecture, we will talk about an application of Gaussian processes to enhance robotic perception. In this context, sensory data (e.g. from depth-sensing cameras or haptic devices) can be used to derive 3D shape information from objects located on a robot’s workspace, which in turn empowers robots to better execute various tasks, such as grasp and motion planning. Finally we will discuss the results obtained from a Master’s thesis on this topic.

Speaker: Gabriela Zarzar Gandler, Master of Science in Machine Learning from KTH (Royal Institute of Technology), is passionate about data science and its applications. The content of the lecture is tightly connected to her Master’s thesis research work. Currently she lives in Stockholm and works as a machine learning research engineer at Looklet.


Watch the recorded presentation on our Youtube channel,



Gabriela Zarzar Gandler

  • Price: Free
  • Organizer: Gabriela Zarzar Gandler
    Logo Gabriela Zarzar Gandler

Det är helt kostnadsfritt att vara medlem på Goto 10.

För att kunna gå på evenemang hos Goto 10 behöver du registrera dig som medlem. Det är helt kostnadsfritt. Vänligen fyll i formuläret nedan.

Obligatoriskt fält
Obligatoriskt fält
Obligatoriskt fält
Obligatoriskt fält

Ange din e-postadress

Ange en giltig e-postadress.