Explore Projects

<p>We add support for sampling arrows to aster models using the theory of curved exponential families.</p>

Accurate modeling of cell temperature is essential for estimating the performance of PV systems. However, very few models exist for calculating cell temperature for floating PV systems, and none of the existing models are available in open-source or commercial software. This project aims to extend pvlib python’s capabilities in modeling floating PV systems by adding functions for calculating cell temperature and ambient conditions. The following will be added to pvlib: - Lindholm model for calculating PV cell temperature - Albedo and wind speed functions - Rahaman thermal model for calculating the PV cell temperature - Gallery example of the expansion of the Faiman expression for floating PV heat loss coefficients Such development will strengthen the pvlib package and enable more accurate modeling of floating PV.

RISC-V Performance Modeling SIG is driving the development of Performance Modeling and Simulation Tools for use across RISCV membership to foster collaboration across the community. To that end, the SIG has developed a trace-driven performance model of an example RISC-V superscalar processor using C++ based on the Sparta simulation framework. The performance model, named Olympia, has been adopted by multiple organizations in both industry and academia. We have identified set of features for further development of Olympia: * Add support for execution-driven modeling - https://github.com/riscv-software-src/riscv-perf-model/issues/14 * Model state of art branch predictors along with decoupled frontend using branch-prediction API, with an example being the BOOM processor frontend documented at https://docs.boom-core.org/en/latest/sections/branch-prediction/index.html - https://github.com/riscv-software-src/riscv-perf-model/issues/143 - https://github.com/riscv-software-src/riscv-perf-model/issues/1 * Develop an API for Interconnect model and implement an example Interconnect microarchitecture. - https://github.com/riscv-software-src/riscv-perf-model/issues/60 * Use the API for data/instruction prefetcher and implement an example prefetcher microarchitecture - https://github.com/riscv-software-src/riscv-perf-model/issues/142 * Enhance the existing vector extension implementation Etc.

<p>In order to make DBpedia and its humongous linked data available to a larger user base in their natural languages (only English for the moment), a Neural QA model has been developed to answer the question in English posed by users. This particular project aims to make our end-to-end system learn better compositionality of questions, by improving our template-generator and our learning model.</p>

This project aims to build a conversational assistant designed around OpenROAD to answer user queries. The goal is to build a complete pipeline that includes dataset updates, model training, and model deployment. The pipeline's architecture will be able to adapt to evolving datasets, with metrics to regularly keep track of the model’s performance.

Bayesian modeling has increased significantly in academia and industry over the past years thanks to the development of high quality and user friendly open source probabilistic programming languages (PPL) in Python and R. Of these is Bambi, a Python library built on top of the PyMC PPL, that makes it easy to specify complex generalized linear multilevel models using a formula notation similar to those found in R. However, as the model building portion of the Bayesian workflow becomes easier, the interpretation of these models has not. Currently, to aid in model interpretability, Bambi only supports conditional adjusted predictions plots. The objective is to take inspiration from the existing Bambi tooling and R package marginaleffects, to extend upon existing plotting functionality and to provide additional plotting functions such as conditional comparisons and conditional marginal effects to allow Bambi modelers to extract insights and interpret their models in a more automatic and effective manner.

<p>In this project, we propose a new package to make graphical models for mixed multi-modal data readily available to a wide audience. The proposed package will allow for fitting, simulating from, and visualizing mixed graphical models. We anticipate that having an easy-to-use R package will increase adoption of these powerful new models.</p>

This project aims to improve the robustness and flexibility of topic modeling in the CHAOSS Augur platform by enhancing its clustering_worker component. Currently, Augur relies on a static number of LDA topics, lacks model versioning, and does not support retraining based on data drift. To address these issues, this project will: Introduce dynamic topic number estimation via Gensim’s HDP model. Implement model versioning with model_id and timestamp support. Record model metadata including hyperparameters and coherence scores. Add automatic parameter tuning using coherence-based grid search. Detect data drift and trigger model retraining accordingly. Export rich visual outputs such as pyLDAvis and wordcloud images. Deliverables include upgraded database schema, a refactored topic modeling pipeline, metadata tables, Jupyter demo notebooks, and HTML visualizations for CHAOSS dashboards. These changes will make Augur’s insight engine more adaptive, interpretable, and future-proof.

<p>codeceptjs-resemblehelper is a CodeceptJS helper which integrates the resemble.js functionality of image comparison in tests used to compare images/screenshots and pass/fail tests based on the tolerance level provided.</p>

Accurately classifying particle collisions—whether distinguishing quark- from gluon-initiated jets or isolating Higgs events from complex backgrounds—requires models that capture both fundamental conservation laws and the inherent geo- metric invariances of the detector environment. This work presents a hybrid self-supervised framework that leverages physics-informed deep learning to ad- dress these challenges. Our approach utilizes a masked transformer encoder to learn conservation of momentum and energy by reconstructing masked particle features, while a Lorentz-Equivariant Geometric Algebra Transformer (L-GATr) encodes the space-time symmetries present in collision events. By integrating these complementary embeddings into task-specific cross-attention decoders, the framework effectively fuses physical and geometric information to enhance event classification and mass reconstruction, thereby advancing the performance and interpretability of machine learning in high-energy physics.

<p>This project will improve the current Appleseed's implementation of physical sun and sky models. Fixing the current blue tint present on Appleseed implementation and adding several new features. Such as, a solar disk, ability to configure the Sun with a geographic location, a date and a time and allow users to change atmosphere parameters. Those new features will make the sky models more flexible for artists and will provide a more accurate representation of the appearance and illumination from the day sky. The new features will also make it possible to render other planets skies, from a physically plausible mars sky to an alien sky with multiple suns and a stunning sky's color.</p>

The Eclipse 4diac project provides an open source infrastructure for distributed Industrial Process Measurement and Control Systems (IPMCS) based on the IEC 61499 standard. This project will add a user interface for testing software components (so-called Function Blocks) based on behavior models. Currently, the tool environment has limited tool support for (semi-)automated testing of Function Blocks. A framework for the test execution is however available and will be used. Expected outcome: A user interface for generating tests based on behavior models and executing them must be available (including documentation). It should be furthermore possible to record new scenarios, which can be later used as test cases.

<p>Implement trajectory transformations on the MDAnalysis API, to be called on-the-fly by the user, eliminating the requirement for multiple intermediate steps of modifying and saving the trajectory, and giving users a more efficient and simple workflow for simulation data analysis.</p>

<p>Static verification of Linux kernel modules has already helped to reveal hundreds of real bugs in Linux kernel device drivers. On practice software verification tools cannot analyze modules of device drivers with the whole Linux kernel source code. Thus verification of separate modules requires generation of a precise and complete environment model to achieve a suitable false positive rate. I propose to develop a set of specifications for an environment model generator of the LDV Tools system developed in terms of the Linux Driver Verification program. It will help to drastically improve verification results reducing the false positive rate and thus will make it possible to find even more real bugs in the latest versions of the Linux kernel.</p>

<p>I aim to implement some important deep learning models in R for the MXNet package. The main component of the proposed work is the implementation of the Recurrent Neural Network (RNN) models and examples. In addition, I will also implement some useful examples in R, such as different image classification models and fully convolutional networks for image segmentation. At the same time, I will write sufficient tests and relevant documentations.</p>

<p>The goal of the project is to use ML techniques to identify relationships between planetary mapped datasets, with the goal of providing deeper understanding of planetary surfaces and to have predictive power for planetary surfaces with incomplete datasets. There are three main proposals for the stated problem:</p> <ol> <li>Explore various ML models, adjust the parameters and compare their performance. During the preliminary work, I already explored several basic ML models; however, we should try more advanced ones and perform a deeper comparison depending on the parameters and metrics.</li> <li>Predict chemicals from chemicals. Based on preliminary work, we see that although in Mercury the correlation between albedo and chemicals is low, the correlationships between some chemicals are pretty high. Thus, in order to fill the "gaps" in the chemical maps, we could try to predict one chemical based on some others. </li> <li>Perform convolutional prediction. Rather than using only one corresponding pixel for the prediction, we could include the surrounding pixels since they might offer additional information.</li> </ol>

This GSoC 2026 project proposes an advanced, open-source computational pipeline for the Late Antiquity Modelling Project (LAMP) to reconstruct historical movement and visual connectivity, moving beyond deterministic archaeological GIS with a data-driven, modular approach. The core objective replaces standard Least-Cost Path analysis with a Probabilistic Path Ensemble. Utilizing Maximum Entropy Inverse Reinforcement Learning (MaxEnt IRL), the pipeline autonomously learns historical cost weights by balancing terrain slope against surface friction derived from K-Means clustered SAR and Multispectral imagery. Additionally, the routing engine leverages spectral anomaly detection and structure tensors to uncover subtle pathways hidden in the landscape. The project's second pillar introduces 3D Viewshed Analysis. By processing high-resolution terrain geometry (.obj files) to calculate precise 3D lines of sight, the toolset enables researchers to evaluate the visual prominence and strategic interconnectivity of ancient settlements and outposts. Built over 12 weeks using robust Python geospatial libraries (Rasterio, GeoPandas, SciPy), the final deliverable is a scalable, mathematically rigorous toolkit. It will serve immediate LAMP needs at sites like El Bagawat while remaining seamlessly transferable to any archaeological context requiring advanced spatial modeling.

<p>This project aims to Develop a Visual Recognition model for Aztec Hieroglyphs. Aztec language is pictographic and ideographic photo-writing. It has no alphabets but different symbolic signs. The model's working will be as follows it'll take in an image, compare it with the ones in our dataset, and finally display the possible matches of that image and identify various atomic elements in the picture. For training, we'll use Deep Neural Network model and, final deployment will be done on Heroku cloud server</p>

Description: WasmEdge would like to build a Rust library crate that enables easy integration of Mediapipe models in WasmEdge applications. Each Mediapipe model has [a description page](https://google.github.io/mediapipe/solutions/face_detection.html) that describes its input and output tensors. The [models](https://google.github.io/mediapipe/solutions/models.html) are available in Tensorflow Lite format, which is supported by the WasmEdge Tensorflow Lite plugin. Expected Outcome: We need at least one set of library functions for each model in Mediapipe. Each library function takes in a media object and returns the inference result.

<p>This project aims to work on two related problems. On one hand, I would like to implement new family of generalized linear models, such as beta regression, robust linear regression (i.e. linear model with error following a T-Student distribution) as well as multinomial regression. On the other hand, it is also necessary to incorporate more automatic prior distribution alternatives that do not make use of the GLM module in statsmodels so new models can work smoothly without requiring the user to manually specify priors.</p>

This project will make the addition and sharing of third-party battery models easier in PyBaMM. It is an extension of the PyBaMM cookiecutter project of year 2024 and will include a Dispatch API, a single location to register models, and enhance model loading using entry points. The copier template will also be enhanced so that individuals with varying levels of experience will be able to use it more easily. By incorporating features such as dynamic model loading, caching, and enhanced documentation, the project will make it easier for more users and developers to work with external models and make PyBaMM more flexible and convenient to use in the long term.

<p>Most appealing for me: Project idea#2: Easier project model management. I am already familiar with customer requirements related to the idea. The idea is interested for me and I have practical experience and knowledge to implement the requirements. So I’ve applied fix for issue #1046 as well as I’ve also implemented on my local development environment fix for issue #1058 . During testing of implemented solution I discovered that related to the fixed issue functionality also can be improved. Detailed calendar plan is described in my Project proposal document.</p> <p>Equally interesting for me: idea 1: Full-featured calendars I am familiar with customer requirements and I know how to implement some of them, other ideas also very interesting for me but require additional investigation to find solution (like an idea #824 Calendars export to other devices.</p> <p>Less interesting for me : idea 3: Search I know how to implement #536 Dashboard project table extra searchable columns. As for second requirement: #535 Full text search engine – this is very interesting task cause it is require a knowledge and experience with Apache Solr. Nevertheless I am ready to take the challenge.</p>

<p>In the R-language, many packages exist for the estimation and forecasting of GARCH processes, including fGarch and rugarch. However, none, to our knowledge, have addressed the issue of robustness toward additive outliers, rather than innovations outliers. Muler and Yohai (2008) proposed two approaches to robust GARCH(p,q) model fitting to avoid bias in the parameter estimates and the preceding kind of over-estimation of volatility following isolated large outlier returns. In the first approach, parameters are obtained by maximum likelihood function using a modified likelihood function based on a bounded loss function. The second approach improves on the first by using a filter that limits the effect of an additive outlier on subsequent predictors of conditional variance. Our proposal is to implement the two approaches by exposing interfaces to a C++ library that can be called from any higher level language for estimating the likelihood functon. We will first develop an R implementation of the robust GARCH(p, q) fitting method, with application examples, and evaluate the performance of the code using as bench- marks selected simulation results from Muler and Yohai.</p>