Find open source projects actively accepting contributors. Search repositories, filter by program milestones, difficulty tags, or tech stack.
Use our Orbit AI Matcher to find out! Get instant matching scores based on your developer skills, preferred frameworks, and contribution experience.
Convert your selected open-source project into a winning GSoC, LFX, or Outreachy application using Proposal Studio.
This project builds an automated benchmarking ecosystem for goal-directed vision. It implements a high-priority leaderboard featuring automatic updates, submission validation, and multi-metric ranking for new Diffusion and VLM paradigms. It ensures reproducibility through versioned results for SOTA research.
Matching papers to reviewers based on topics is a crucial task for the Neurons, Behavior, Data Analysis, and Theory (NBDT) journal. However, the current automatic reviewer assignment tool that uses SciBERT embeddings, cosine similarity, and linear programming may not capture the semantic meaning of the text accurately. This project aims to address this limitation by finetuning SciBERT on the relevant corpus of data and selecting the appropriate optimization objectives. SciBERT can learn from both the left and right contexts of words and has a vocabulary that is more suitable for scientific texts than BERT. The project will involve creating a training dataset, pre-processing it, generating the appropriate features, selecting and fine-tuning SciBERT, generating the word embeddings from the fine-tuned SciBERT and choosing the appropriate optimization objectives (Contrastive Learning, Learning to Rank Diversely, and LambdaRank) according to the dataset obtained and the subsequent evaluation of its performance. The expected outcome is an improved tool that more accurately matches papers to reviewers for the NBDT journal and can potentially be useful in other domains as well.
<p>PEcAn is an ecological modelling toolbox with the capability to automate, preserve, and repeat model runs. One functionality still in the early stages is benchmarking, or comparing models to empirical data. My project aims to add data ingest and cloning functionality to the benchmarking module of PEcAn to make it applicable to a wider range of models and regions and more robust in its ability to benchmark multiple models.</p>
Acquiring neural fMRI and abdominal MRI data is inherently expensive and time-consuming, significantly limiting the volume of datasets available for advanced analysis. Consequently, the full potential of GI tract disease diagnosis and segmentation models remains unrealized, as limited datasets hinder their accuracy and generalizability. Diffusion models offer a promising solution to these challenges. By effectively learning intricate spatiotemporal patterns inherent in fMRI/MRI data, diffusion models can synthesize high-quality, realistic data samples. This capability augments available datasets, substantially enhancing the accuracy and reliability of GI tract semantic segmentation and disease diagnosis models.
This project involves creating a VS Code extension that provides enhanced support for the Brian2 neural simulator. The extension aims to improve the user experience and productivity of those working with Brian2 by offering syntax highlighting, syntax checking, auto-completion, and language server protocol (LSP) support. These features will be implemented using various programming languages and tools, including Javascript, Python, and the LSP protocol. The deliverables for this project include a fully functional extension that can be installed and used with VS Code, as well as documentation on how to install and use the extension. The extension will be designed to work seamlessly with Brian 2 and will provide a user-friendly interface that is easy to use and navigate.
This project aims to create a high-quality custom dataset using PaLM API and create a high-quality text summarization model by fine-tuning a pre-trained model with the custom dataset. The custom dataset will be a CSV file with 10K rows and 4 columns: id, prompt, document, and summary. The document should vary in different writing styles such as news articles, blog posts, legal papers and so on, and the length of a summary will be between 1/3 and 1/4 of the length of the original document. This will make the custom dataset to be far superior to most publicly available datasets like CNN-Daily-Mail or XSUM, which typically generate 1-2 sentence summaries. Then I will create a few text summarization models with different pre-trained models, improving each fine-tuned model’s accuracy by implementing strategy driven from bias/variance and error analysis. After creating a few fine-tuned models, I will pick the best performing one based on the evaluation metric. Based on this, I will work on creating a tutorial covering the whole process, such as creating a custom dataset using PaLM API, fine-tuning pre-trained models using Keras API, and selecting the best performing one based on a evaluation metric. As a result, by creating this comprehensive tutorial, the project will encourage other developers to explore/use Keras API and PaLM API in their projects.
<p>As Lütkepohl [1] states his preface, "[b]y now [cointegration] related models and methods have become far more important for applied econometric work than, for example, vector autoregressive moving average (VARMA) models." In the context of cointegrated variables especially vector error correction models (VECM) are emphasized in [1].</p> <p>While statsmodels supports VARMA models, it still lacks VECM (<a href="https://groups.google.com/d/msg/pystatsmodels/Dr3_c4BMaiY/JYx0H0gsBgAJ" target="_blank">as suggested by one of the project's main contributors</a>). The GSoC might be a good opportunity to introduce VECM-related functionality into statsmodels. This functionality includes the representation of a VECM, forecasting, estimation, and model diagnostics among other topics.</p> <p>[1] Lütkepohl (2005) New Introduction to Multiple Time Series Analysis</p>
<p>A recent addition to the local statistical models in PySAL is the implementation of Multiscale Geographically Weighted Regression (MGWR) model, a multiscale extension to the widely used approach for modeling process spatial heterogeneity - Geographically Weighted Regression (GWR). The GWR model in PySAL can currently estimate Gaussian, Poisson and Logistic models though the MGWR model is currently limited to only Gaussian models. This project aims to expand the MGWR model to nonlinear local spatial regression modeling techniques where the response outcomes may be discrete (following a Poisson distribution) or binary (Logistic models). Subsequently, to support efficient testing for different model implementations, a simulated data generator module will be implemented to supply test datasets following unique model variable distribution needs. This will also provide a foundation for possible expansion to test other local model implementations in PySAL. Additionally, since the functionality to predict the dependent variable at unsampled locations is not supported for the MGWR model in PySAL, this project also aims to enable predictions for MGWR.</p>
<p>Currently, in the Splash there is no an easy way to click a link, fill and submit a form, extract data from a webpage using Splash Scripts. I will develop a helper library with some helper classes like Element which will represent DOM Element. This class will allow to interact with DOM elements using Lua without any need of using splash#evaljs function.</p>
<p>The current automatic classification system in place for the ptwiki is very naive and simply checks a few if conditions and places articles accordingly. There are 6 target labels that the existing system places articles into 2 of which require editor approval. The proposal elaborates on an implementation of articlequality for ptwiki similar to those deployed for other wikis.</p> <p>Such an implementation would require feature extraction from ptwiki, training various models on these features and fitness testing these models to find the best fit.</p> <p>The immediate use cases of the articlequality model would be:</p> <ol> <li>Help increase the quality of automated article classification for ptwiki</li> <li>Streamline work for editors on ptwiki with respect to finalizing articles that need expansion, improvements or articles that can be featured.</li> </ol> <p>The draftquality model will help eliminate vandal drafts faster and save patroller time.</p> <p>The implementation would also pave the way for further work to be done in automating various wiki tasks for ptwiki.</p> <p>The proposal also hopes to apply insights learned from deploying the articlequality and draftquality models for ptwiki to the uk, bs and hi wikis.</p>
<p>Diffusion MRI measures water diffusion in biological tissue, which can be used to probe its microstructure. The most common model for water diffusion in tissue is the diffusion tensor (DT), which assumes a Gaussian distribution. This assumption of Gaussian diffusion oversimplifies the diffusive behavior of water in complex media, and is known experimentally to break down for relatively large b-values. DT derived indices, such as mean diffusivity or fractional anisotropy, can correlate with major tissue damage, but lack sensitivity and specificity to subtle pathological changes. Microstructure Imaging of Crossing (MIX) is versatile and thus suitable to a broad range of generic multicompartment models, in particular for brain areas where axonal pathways cross.</p> <p>Multicompartment models (assess the variability of diffusion in sub-voxel regions) enable the estimation of more specific indices and so potentially give much greater insight into tissue architecture.</p> <p>Goal of Model Fitting: Identify which model compartments are essential to explain the data and parameters that are potentially estimable from a particular experiment and compare the models to each other using the BIC.</p>
<p>Speech is a complex phenomenon. People rarely understand how is it produced and perceived. The naive perception is often that speech is built with words, and each word consists of phones. The reality is unfortunately very different. Speech is a dynamic process without clearly distinguished parts. It's always useful to get a sound editor and look into the recording of the speech and listen to it. All modern descriptions of speech are to some degree probabilistic. That means that there are no certain boundaries between units, or between words. Speech to text translation and other applications of speech are never 100% correct. That idea is rather unusual for software developers, who usually work with deterministic systems. And it creates a lot of issues specific only to speech technology. The system is intended to develop Language model and Acoustic model to be used for speech recognition of Malayalam Language.</p>
<p>During the summer, I am going to implement Kim Filter, which is broadly used for regime switching state-space models evaluation, on top of existing Kalman Filter. I will also create a bunch of well-tested, ready-to-use models with regime switching, including MS-AR, MS-DFM and MS-TVP.</p>
<p>During the project the current model construction will be analysed in regard to performance(i.e. reachability of multiple initial states, scalability and comprehensiveness(adding support for other types of models, i.e. PTAs). Depending on the analysis, it is decided which changes in the current approach enhance the functionality/ performance/ scalability and together with the supervisor a feature list can be defined. Then the implementation of the features is conducted and additionally efficient transformation to a sparse model and full import/export of the explicit model is implemented.</p>
<p>ASCEND provides support for Conditional modeling, where a model can contain both variant and invariant equations. Invariant equations must be respected by the model at all times while the variant equations only are included in the model if certain conditions are met. For example, a model for fluid flow will have different sets of equations depending on the phase of the flow (laminar, turbulent or sonic flow). Unfortunately, the current syntax for conditional modeling has been considered unintuitive by some of its users and there is demand for a better alternative.</p> <p>In this project, my goal is to precisely identify why the current syntax is considered unintuitive and use this information to design and implement a better alternative. The identification would be based on a review of code examples using the current conditional model syntax and the implementations would be made by modifying the existing Bison-based compiler.</p>
My proposal aims to optimize PyTorch Image Models (Timm) using OpenVINO and NNCF libraries. The problem at hand is that these models are computationally intensive and require a lot of resources, making them challenging to deploy in resource-limited environments. To solve this problem, I plan to implement optimization techniques such as quantization, pruning, and compression on a selected subset of models from the Timm library. I will evaluate the performance of these optimized models on different hardware platforms and compare them with the baseline models. The deliverables for this project include optimized Timm models, a tutorial notebook detailing the setup and usage of OpenVINO and NNCF libraries, and documentation and tests for the code developed during the project. If my proposal is accepted, I am excited to work on this project and contribute to the community by making these models more accessible and efficient to deploy.
<p>The new implementation will consist of a central registry in <code>ProjectState</code><em> instances where all the related fields would be registered for all apps. Then significant changes in <code>ProjectState</code></em>, Schema Editor, and <code>Operations</code><em>' subclasses would be made in order to make all of them adapt ModelStates in spite of fake rendered models in a backward-compatible way. Then the new method mapping in <code>ProjectState</code></em> for all Operation subclasses' <code>state_forwards</code> and <code>state_backwards</code> would be introduced. And after all this, the traditional <code>__fake__</code> model implementation would now be replaced with new <code>ModelState</code> implementation</p>
<p><strong>Improve the Ersilia Model Hub, a FOSS platform offering pretrained AI/ML models for research</strong></p><p>The deadline to create an initial application has passed for the May 2022 Outreachy internship cohort. We are no longer accepting initial applications for internships. We encourage you to sign up for the announcements mailing list to get an email when the next round opens. Why apply to Outreachy? > Start your initial application > May 2022 Outreachy internship cohort > Ersilia Open Source Initiative Community details are hidden until you are approved to participate as a mentor or coordinator.</p><p><br></p><p><strong>Mentorship Cohort:</strong> 2022</p>
<p><strong>Extension of the Ersilia Model Hub, a FOSS platform featuring AI/ML models for biomedical research</strong></p><p>The deadline to create an initial application has passed for the December 2022 Outreachy internship cohort. We are no longer accepting initial applications for internships. We encourage you to sign up for the announcements mailing list to get an email when the next round opens. Why apply to Outreachy? > Start your initial application > December 2022 Outreachy internship cohort > Ersilia Open Source Initiative Community details are hidden until you are approved to participate as a mentor or coordinator.</p><p><br></p><p><strong>Mentorship Cohort:</strong> 2022</p>
Developing a scalable and reproducible research pipeline within SeqTrainer for bacterial DNA sequence modelling and annotation using SBOL-based datasets. The project addresses a key challenge in synthetic biology: the lack of accessible, machine learning–ready datasets and standardized workflows for genomic modelling. The primary focus is on E. coli and Gram-negative bacteria, with applications in promoter classification and sequence annotation. The work involves reproducing and benchmarking existing CNN-based baselines, followed by integrating and evaluating foundation models such as DNABERT2 and Evo 2 within structured experiment pipelines. Experiments explore challenges including overfitting and class imbalance through methods such as weighted loss functions, undersampling, learning-rate tuning, and early stopping. A major component of the project is the development of a modular and reproducible experiment framework within SeqTrainer, including configuration-driven workflows, reproducible dataset splitting, evaluation utilities, metric logging, and improved SBOL-to-dataset conversion pipelines with validation and provenance tracking. The final outcome aims to transform SeqTrainer into a more experiment-friendly framework for genomic ML research, enabling accessible, extensible, and reproducible machine learning workflows for synthetic biology researchers.
<p>The aim of the project is to facilitate NLP tasks in DFFML by adding support for two of the most popular NLP libraries: HuggingFace Transformers and spaCy.</p>
<p>The Energy Hub (Ehub) Modeling Tool is a set of MATLAB scripts for writing and executing a district energy system optimization. The purpose of the tool is to aid the design of decentralized energy systems at the building and district level. It can be used for creating, executing and visualizing the results of an energy hub model for a given case study and a set of technologies. The Ehub Modeling Tool takes as input case study data and technology data in defined formats and outputs an AMS file which is executed in the optimization package Aimms. The Ehub Modeling Tool also includes R code for visualizing the results of the energy hub model.</p> <p>This project focuses on developing a Python-based visualization module for the Ehub Modeling Tool. The results from a run of the model include the optimal dispatch schedule of different technologies, the technologies to be installed and the optimal capacities of these technologies, and other data. This project involves developing a set of scripts to import and visualize the results data from a set of experiments.</p>
<p>This project intends to define user interface to Customize Model interactively instead of modifying them manually. It will make the user able to create the templates for given model which can further be changed as per user’s requirements.</p>
<p>Computational models of biological systems play an important role in advancing science as they give insights about the biology without the need for performing real experiments. A significant challenge in the systems biology field is the reproducibility and reusability of such models and making these accessible to the broader biology community. To address this issue, computer-readable formats and tools for modeling complex biological systems have been developed. Standards such as Simulation Experiment Description Markup Language (SED-ML) and tools like Systems Biology Simulation Core Library (SBSCL) can represent such complex biological models and numerically simulate them. But until now only a subset of these standards have been implemented in SBSCL. Essential features that will catalyze reproducibility, exchangeability, and reusability are lacking so far. Namely, i) implementation of full support for simulation experiments; ii) plots and visualization of simulation studies; iii) support for standard archive formats; iv) support for Open Source solvers, and v) support for hierarchical models.</p>