Group of Machine Learning Research
Location: Kraków, Poland Website: gmum.net Available projects: 35
GMUM (Group of Machine Learning Research) is a group of researchers working on various aspects of machine learning, and in particular deep learning - in both fundamental and applied settings. The group is led by prof. Jacek Tabor. We are based in the Jagiellonian University in the beautiful city of Kraków, Poland.
Some of the research directions our group pursues include:
- Generative models: efficient training and sampling; inpainting; super-resolution,
- Theoretical understanding of deep learning and optimization,
- Natural language processing,
- Drug design and cheminformatics,
- Unsupervised learning and clustering,
- Computer vision and medical image analysis.
In 2023, we organized the second edition of Machine Learning Summer School (MLSS^S) with a focus on Applications in Science. We invite participants to collaborate with us on various ongoing research projects - learn more here.
Lab's projects:
The project aims to create a unified benchmark for multiple methods that reduce the inference time of deep learning models. We begin by focusing on early-exiting methods. You task will be to reimplement a conditional computation method from a selected published paper into our common codebase. Conditional computation methods are usually simple to implement and provide significant computational cost savings. We intend to publish the benchmark with the accompanying analysis as a paper in a rank A* conference. We have tasks appropriate for both beginners and people with experience.
General must-have requirements
The student needs to know how to program in PyTorch (In order to undestand the present code and implement new methods).
Contact: Bartosz Wójcik, Maciej Wołczyk (bartwojc [ at ] gmail.com)
See project's pageYou will simulate two-sided urban mobility market (like Uber or Lyft), where agents get rewarded for their actions. In particular travellers can decide among platforms (Uber or Lyft) or opt-out (use public transport) - based on prevoious experiences. They, however, need to learn which actions are otpimal for them (subjectively). You will use https://github.com/RafalKucharskiPK/MaaSSim and apply decision modules to the agents.
General must-have requirements
PyTorch or Tensorflow
Contact: Rafał Kucharski (rafal.kucharski [ at ] uj.edu.pl)
See project's pageSGD optimization is currently dominated by 1st order methods like Adam. Augumenting them with 2nd order information would suggest e.g. optimal step size. Such online parabola model can be maintained nearly for free by extracting linear trends of the gradient sequence (arXiv: 1907.07063), and is planned to be included for improving standard methods like Adam.
General must-have requirements
The students needs to know basics of tensor flow or pytorch, preferred experience in mathematical analysis.
Contact: Jarek Duda (jaroslaw.duda [ at ] uj.edu.pl)
See project's pageThe aim of this project is to create a new Normalizing Flow models in a style of diffusion ore autoregressive models.
General must-have requirements
PyTorch
Contact: Marcin Sendera (marcin.sendera [ at ] gmail.com)
See project's pageRecently my academic focus is on continual learning. I am interested in Bayesian learning, optimization. Any reasonable combination (potentially with generative modelling) would be a fantastic project for me. If you don’t have any particular ideas, I do have something to offer!
General must-have requirements
Nice to have strong maths background.The student needs to know how to program in either tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Mateusz Pyla (mateusz.pyla [ at ] doctoral.uj.edu.pl.)
See project's pageThe objective is to enable control over the process of generating objects in models like StyleGAN (and others). A solution has already been implemented, as detailed here: https://ojs.aaai.org/index.php/AAAI/article/view/20843. In addition to developing these types of models, we aim to use them for generating counterfactual examples, which facilitate the explanation of machine learning model predictions.
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Marek �mieja (marek.smieja [ at ] ii.uj.edu.pl)
See project's pageIn continual learning we want to understand the phenomenon of catastrophic forgetting - network quickly losing performance at previously learned tasks after encountering new tasks. However, this usually concerns zero-shot forgetting – what happens if we’re allowed to quickly recall the old problem before attempting to solve it? The goal of the project is to investigate how quickly we can recall the “forgotten” knowledge and build a CL method optimized for that
General must-have requirements
PyTorch or Jax
Contact: Maciej Wołczyk (maciej.wolczyk [ at ] gmail.com)
See project's pageThe self-supervised learning model allows for constructing a data representation in an unsupervised manner, which can later be successfully used for classification or clustering. However, much depends on the augmentations used. If an augmentation changes the class, it becomes difficult to utilize such a representation in classification later. In this project, we aim to build a model that constructs a representation that is less sensitive to the type of augmentations used.
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Marek Śmieja (marek.smieja [ at ] ii.uj.edu.pl)
See project's pageTwo interesting projects:
- Implicit representation of diffusion - the goal is to reduce the architecture so that it can operate on standard cards.
- Diffusion on network weights.
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Przemysław Spurek (przemyslaw.spurek [ at ] uj.edu.pl.)
See project's pageYou will simulate two-sided urban mobility market (like Uber or Lyft), where agents get rewarded for their actions. In particular drivers (or Connected Autonomous Vehicles- CAV) can reposition and wait for requests in different part of the city. They, however, need to learn when and where it is efficient to reposition. You will use https://github.com/RafalKucharskiPK/MaaSSim and apply decision modules to the agents.
General must-have requirements
PyTorch or Tensorflow
Contact: Rafał Kucharski (rafal.kucharski [ at ] uj.edu.pl)
See project's pageBased on an interpretable backbone [either b-cos or prototypes]. We have a proposed tree model that allows for storing the structure of non-binary trees.
General must-have requirements
PyTorch
Contact: Jacek Tabor, Łukasz Struski (jacek.tabor [ at ] uj.edu.pl lukasz.struski [ at ] uj.edu.pl)
See project's pageWe want to extend our work on early-exiting to also accelerate the training process.
General must-have requirements
PyTorch
Contact: Bartosz Wójcik (bartwojc [ at ] gmail.com)
See project's pageThe goal of the project is to build a single network (using the concept of a hypernetwork) that can generate various networks to solve a single task.
General must-have requirements
Pytorch
Contact: Jacek Tabor, Przemysław Spurek (jacek.tabor [ at ] uj.edu.pl)
See project's pageThe aim of this project is to extend the Non-Gaussian Gaussian Processes framework in the sense of flexibility (e.g., adding the conditional case based on support set data).
General must-have requirements
PyTorch
Contact: Marcin Sendera, Tomasz Kuśmierczyk (marcin.sendera [ at ] gmail.com)
See project's pageWe are looking to extend our Continual World benchmark: https://arxiv.org/abs/2105.10919 in various ways, such as learning from pixels, implementing new RL algorithms, implementing new continual learning methods, exploring sparse rewards setting.
General must-have requirements
Python, preferably TensorFlow 2
Contact: Maciej Wołczyk (maciej.wolczyk[at]gmail.com)
See project's pageFew-shot learning (FSL), also referred to as low-shot learning (LSL) in few sources, is a type of machine learning problems where the training dataset contains limited information. Few-shot learning aims for Deep learning models to predict the correct class of instances when a small amount of examples are available in the training dataset.
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Przemysław Spurek (przemyslaw.spurek [ at ] uj.edu.pl)
See project's pageIn the project we will use GAN for generating NeRF reprezentaions NeRF https://www.matthewtancik.com/nerf
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Przemysław Spurek (przemyslaw.spurek [ at ] uj.edu.pl.)
See project's pageWe want to find a way to generate chemical molecules that is useful from the perspective of drug design process. Primarily, the new generative model should be able to follow given structural constraints and generate structural analogs, i.e. molecules similar to previously seen promising compounds.
General must-have requirements
PyTorch and Tensorflow
Contact: Tomasz Danel, Łukasz Maziarka (tomasz.danel [ at ] ii.uj.edu.pl)
See project's pageThe goal of the project is to generate high-quality models of human faces using the NeRF algorithm.
General must-have requirements
PyTorch
Contact: Przemysław Spurek (przemyslaw.spurek [ at ] uj.edu.pl.)
See project's pageNeural networks achieve very good results in typical classification and clustering tasks. However, the interpretation of the obtained results is limited. In this project, we aim to focus on constructing deep models that make predictions by undertaking a sequence of decisions.
Intuitively, we are dealing with models that build a decision tree/graph, allowing for an enhanced interpretation of the results. A model has already been implemented and detailed in the https://arxiv.org/abs/2107.13214.
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Marek Śmieja (marek.smieja [ at ] ii.uj.edu.pl)
See project's pageFoundation models have delivered impressive outcomes in areas like computer vision and language processing, but not as much in reinforcement learning. It has been demonstrated that fine-tuning on compositional tasks, where certain aspects of the environment may only be revealed after extensive training, is susceptible to catastrophic forgetting. In this situation, a pre-trained model may lose valuable knowledge before encountering parts of the state space that it can handle. The goal of the project is to research and develop methods which could prevent forgetting of the pretrained weights and therefore get better performace by leveraging previous knowledge. Highly recommend section 4.4 Minecraft RL
General must-have requirements
PyTorch
Contact: Maciej Wołczyk, Bartłomiej Cupiał (maciej.wolczyk [ at ] gmail.com)
See project's pageAn open-source library for transformer-based molecular property prediction with a simple and unified API that provides the implementation of several state-of-the-art transformers for molecular property prediction. The library is in the development stage, and there are many interesting things to be implemented: novel transformer-based models, pre-training methods, integration with huggingface caching system, Continous Integration, and a few other things. The complexity of the tasks is diverse, ranging from “good first issue” to “game-changer”, so basically, anyone can find something suitable :)
General must-have requirements
PyTorch
Contact: Piotr Gaiński, Łukasz Maziarka, Tomasz Danel i Stanisław Jastrzębski (piotr.gainski [ at ] student.uj.edu.pl)
See project's pageNeural networks achieve very good results in popular domains such as images or texts. However, for tabular data, which does not possess a local structure, shallow methods like random forest or XGBoost often achieve better results. The goal is to develop neural network models that can be successfully applied to tabular data.
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Marek Śmieja (marek.smieja [ at ] ii.uj.edu.pl)
See project's pageThe objective is to develop a MIL (Multiple Instance Learning) model that is designed to be memory efficient. We plan to base our approach on the study detailed in this https://proceedings.mlr.press/v202/zhu23l/zhu23l.pdf.
General must-have requirements
PyTorch
Contact: Jacek Tabor, Łukasz Struski (jacek.tabor [ at ] uj.edu.pl lukasz.struski [ at ] uj.edu.pl)
See project's pageFor the echocardiography (USG of the heart) [there are two different projections or snippets of film in the same class]. We want the model to build a representation that is independent of the choice of projection.
General must-have requirements
PyTorch
Contact: Jacek Tabor, Łukasz Struski (jacek.tabor [ at ] uj.edu.pl lukasz.struski [ at ] uj.edu.pl)
See project's pageIn the project, we aim to build models that allow for generating chemical molecules. Specifically, we want the generated molecules to satisfy certain conditions, such as activity, solubility, the number of rings, etc.
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Marek Śmieja (marek.smieja [ at ] ii.uj.edu.pl)
See project's pageIn the project we will use GAN for generating NeRF reprezentaions NeRF https://www.matthewtancik.com/nerf
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Przemysław Spurek (przemyslaw.spurek [ at ] uj.edu.pl.)
See project's pageThe aim of this project is to utilize the Normalizing Flows and other Generative models for architectures used for very large Meta-Learning datasets.
General must-have requirements
PyTorch, Tensorflow
Contact: Marcin Sendera (marcin.sendera [ at ] gmail.com)
See project's pageInterpretation of the last convolutional layer in convolutional models. We want to create a segmentation model that is interpretable similarly to GradCam, but one that possesses information allowing for correct classification.
General must-have requirements
PyTorch
Contact: Jacek Tabor, Łukasz Struski (jacek.tabor [ at ] uj.edu.pl lukasz.struski [ at ] uj.edu.pl)
See project's pageIn the dataset of 5mln trips made with Uber in Chicago some of the are pooled - travel together (20%). Which and why. Can we use this dataset to sucesfully predict which of them will be pooled and what factors influence it? This paper scratched the surface, let’s go deeper: https://doi.org/10.1177/0361198120915886
General must-have requirements
PyTorch or Tensorflow, pandas, XAI
Contact: Rafał Kucharski (rafal.kucharski [ at ] uj.edu.pl)
See project's pageWe are interested in developing a https://dl.acm.org/doi/pdf/10.1145/3447548.3467245 that will enable detailed video analysis, potentially incorporating https://arxiv.org/pdf/2301.12276.
General must-have requirements
PyTorch
Contact: Jacek Tabor, Łukasz Struski (jacek.tabor [ at ] uj.edu.pl lukasz.struski [ at ] uj.edu.pl)
See project's pageYou will apply ride-pooling algorithm which pools travellers (e.g. of Uber) into attractive groups. You will use ExMAS (https://github.com/RafalKucharskiPK/ExMAS) which provides exact analytical search in the combinatorically exploding search space (e.g. for 1000 trip requests there is almost a googol number of possible 5-person groups). You will use this analytical results to train supervised machine learning and explore the ways to make the search space searchable.
General must-have requirements
PyTorch or Tensorflow, optimization, ILP, networkX
Contact: Rafał Kucharski (rafal.kucharski [ at ] uj.edu.pl)
See project's pageWhat if not all objects on single sample are labeled? Can we develop method for learnig deep models in such case?
General must-have requirements
The students needs to know basics of tensor flow or pytorch, preferred experience in mathematical analysis.
Contact: Krzysztof Misztal (krzysztof.misztal [ at ] uj.edu.pl)
See project's pageApplying a transformer model to something akin to U-NET that segments [it is supposed to learn background removal in an unsupervised manner].
General must-have requirements
PyTorch
Contact: Jacek Tabor, Łukasz Struski (jacek.tabor [ at ] uj.edu.pl lukasz.struski [ at ] uj.edu.pl)
See project's pageThis project is centered around representing sound using a network.
General must-have requirements
The student needs to know how to program in both tensorflow and pytorch (In order to undestand the present code and implement new methods).
Contact: Przemysław Spurek (przemyslaw.spurek [ at ] uj.edu.pl.)
See project's page