ML4Science

Interdisciplinary Machine Learning Projects Across Campus

As part of the Machine Learning Course CS-433, students can bring their ML skills to practice by joining forces with any lab on the EPFL campus, or other academic institutions, across any disciplines and topics of their choice.

In the four editions so far, 310 collaborative projects have been successfully completed, across 105 different labs of EPFL and other institutions.

The project is done in a group of 3, and counts 30% to the grade of the course. Detailed project description.

Call for projects: If your lab (from any academic institution in CH) wants to host a group of student for the next iteration of the course, please fill the contact form here.

Past projects: The interdisciplinary ‘ML4Science’ projects performed last years across campus were (including reproducibility challenge):

2023:
ML4Science:

• Image-based Cell Cycle Phase Classification using Machine-Learning Techniques
• Learning phase masks for lensless imaging
• Swiss Plasma Center: Investigation of the operational parameters for QCE H-mode regime access in TCV plasmas with supervised learning
• Deep learning image classification for cell cycle stages
• Measuring road network expansion and its effect
• Recovering HDR signals from LDR images for understanding contrast preservation for view-out research
• Data-driven surrogate modeling of the hemodynamics by prediction of POD coefficients (3D Navier-Stokes equations)
• Advancing Homepage2Vec with LLM-Generated Datasets for Multilingual Website Classification
• BERT-based Synthetic Time Series Generation for Classification Tasks
• Data-driven Marketing Strategies for Non-Profit Organizations
• Reinforcement Learning Actor for Blade Pitch Control in Wind Turbine Systems
• Classification of Hsp70 proteins into taxonomic groups based on their amino acid sequences
• Predicting mortgage default in the US
• Synthetic data generation with VAE for fairness mitigation
• ModelAgnostic Meta-Learning for Drug Response Prediction
• Fine-tuning open-source LLMs for argument detection
• Packed Ensemble Surrogate Models for Fluid Flow Estimation Arround Airfoil Geometries
• Urban Heat Island intensity prediction using ensemble regression methods
• IPCA with RKHS Ridge Regression
• ML methods to accelerate molecular dynamics simulation
• Protein Thermal Stability Prediction
• Reproduction and Adaptation of Epileptic Seizure Detection Algorithm into Standardization Framework
• Investigation of the operational parameters for QCE H-mode regime access in TCV plasmas with supervised learning
• Explaining venture teams’ opportunity identification through multimodal data
• Knowledge Tracing: Comparing DNNs Versus GPT For Intelligent Tutoring Systems
• Commonsense Persona Grounded Dialogue challenge
• Optimizing Beam Dynamics in LHC
• Advancing Medical Diagnosis: A Comparative Study of Machine Learning Models and Text Embedding Techniques
• SCRAP: Seismic collapse response prediction of steel moment resisting frames
• Cosmic Reionization Simulations using PINNs
• Testing the McGurk Effect on Multimodal Audiovisual Models
• Categorising Buildings in Lausanne Based on Facade Material
• Analysis of View-out Impression Data
• Image processing/pattern recognition on MHD spectrograms to automate the detection of phase in the discharge characterized by Magneto-HydroDynamic instabilities
• EEG decoding
• Performing 2D Nucleus Segmentation With Cellpose and StarDist on Xenium Data
• Meditron-V: Empowering Meditron with Visual Understanding
• Recurrent convolutional neural network for coronal jets identification
• Project 2 Submission – Laurel and Hardy
• Estimate vocabulary size of young second language learners based on speech-to-text transcripts
• ML-powered Tool for assisted design of Buildings
• ML4Science in collaboration with LAPD
• MEDIAR
• Machine Learning for Chess Movement Recognition
• Neural Control Systems (ML4S, reproduction paper)
• ML4Science Project: Exploration for mutants with enhanced protein fitness
• Crafting Metrics for MiniMeditron Reward Models
• Coronal Mass Ejection prediction from spacecraft EDAC measures
• Enhancing Fear Detection in GBV Survivors: A Deep Learning Approach with Physiological Data Analysis
• Identification of radiative instability patterns
• Predicting lipid abundance in a murine brain section from spatial gene expression
• ML4science: Protein fitness landscape & optogenetic design
• Generalist neural networks for segmenting microscopy images
• Machine Learning for chess moves recognition
• State space modeling of memory representations in the hippocampus
• Data-driven surrogate modeling of the hemodynamics by prediction of POD coefficients
• Explaining venture teams’ opportunity identification through multimodal data
• Commonsense Persona-Grounded Dialogue Challenge (CPDC 2023)
• Machine Learning Based Spherical Deconvolution for Intra-Voxel Fiber Estimation and Brain Connectivity Mapping
• Efficiency of BlazePose markless pose estimation
• Exploring Brain Modulation: Opto-fMRI Decoding of Locus Coeruleus Firing Patterns Using Searchlight, Multiple Kernel Learning, and Graph Neural Networks
• Exploring Neural Scaling Law for Deep Neural Network Models of the Ventral Visual Stream
• Deep Biometric Fingerprint Matching
• Advancing R&D Expenditure Predictions: The Power of Google Trends Data in Nowcasting
• Physiotherapy Exercises and Errors Classification
• Machine Learning Force Fields (ML-FFs) from Spatial Equivariant Descriptors
• Benchmarking models for cell location recovery from single cell transcriptomics
• LLM Personalities Shenanigans
• Biofilll: Evaluating Extreme Reconstruction and Inpainting on Biological Imaging
• Topological analysis of Drosophila neural cultures
• Comparisons and Ratings in a Recommender System
• Machine Learning for Science: Clustering of DNA Sequences for Primer Extraction
• Beyond additive noise: Evaluating Style Geometric Style Transfer in Scene Text Recognition
• Characterization and automatic differentiation between minor and major disruptions
• Automated Informalization Using LLemma
• Markerless Pose Estimation
• ML4S – Train generalist deep neural networks with database of annotated images
• A Dual-Task Pipeline for Predicting Optic Nerve Neuron Firing Patterns
• Optimisation of question selectionfor long term memory retention using machine learning
• Rethinking Optimization for Non-Convex problems with SCRN and its derivatives
• Enhancing Home Physiotherapy: Machine Learning for Exercise Recognition and Error Detection
• Lesion segmentation for Magnetic Resonance Images (MRI)
• CME detection onboard Venus Express
• Forecast the diagnostic group (PTSD or CUD) from the intrusion characteristics
• CERN Compressor Vibration Analysis
• Generative AI for Energy Retrofits
• LatentVec: Evaluating Generative Models for Vector Graphics with Interpretable latent states.
• Predicting Patent Success: A Machine Learning Approach to Commercialization Prospects
• Using LLMs for Retrieval Augmented Generation on the RTS Archive
• ML4Science project : Identifying the neural predictors of subjective sleep quality from sleep EEG recordings in trans-diagnostic psychiatric conditions
• Multi-category Multi-object Indoor Scenes Understanding for Intelligent Assistive Environment
• Radiated power reconstruction given BOLO/AXUV line integrated measurements
• Predicting brain activity during movie watching: from the movie frames to the brain
• Classification Using Natural Gradient Descent
• Movement recognition for physiotherapy
• Generalizable Yeast Cell Segmentation
• Investigation of the operational parameters for QCE H-mode regime access in TCV plasmas through unsupervised learning
• Detecting shapes of nanoparticles in TEM-images
• Fine-tuning LLMs for argument quality assessment
• Novelty analysis of business ideas proposed by startup teams from various industries
• Exploring Transformer Models for Page Prefetching
• Fine-tuning Large Language Models for Argument Stance Detection in Unseen Domains
• Does making a vision model more brain-like make it more robust?
• Estimation of Brazil Nut Productivity in the Amazon (Madre de Dios)
• Train generalist deep neural networks with database of annotated images: MEDIAR Architecture
• Reproduction of Vehicle Trajectory Prediction with an LSTM-based Kalman Filter
• Federated Learning with Partial Model Personalization

2022:
ML4Science:

• Predicting poverty through time with publicly available data
• Multimodal Deep Learning for a Global data set of COVID-19 Patients
• Classification of Tweets based on Censorship
• Predicting Myocardial Infarction from In Silico Data Using Transfer Learning
• Fine-tuning and Prompt-learning on Commonsense Causal Reasoning
• Modelling Energetic Particles in Matter
• Predicting poverty through time with publicly available data
• Improving 3D cell segmentation in whole-brain imaging
• Multitask learning for harmodynamics
• Autonomous Lane Changing using Deep Reinforcement Learning with Graph Neural Networks
• Optimal U-Net for Image Segmentation
• Automatic segmentation of nerve histological section – TNE lab
• Solubility classification of molecules
• Censorship of Twitter – Unsupervised Topic Modeling
• StockDataHolm
• Classify protein molecular surfaces from their interaction fingerprints
• Using Machine Learning for the Prediction of Motor Impairment and Recovery in Stroke
• Fear classification from physiological information using knowledge distillation
• Probe Posture Prediction
• Automated Inpaintings of Humans in Images
• Shedding Light on Mouse Sleep: Automated Sleep Phase Detection using Machine Learning
• Predicting the Intelligence of a Person From a Video: A Deep Learning Approach
• SimpledCLIP: An Attempt to Recover Prompted Human 3D Poses using CLIP
• Deep Learning for Edge-AI Respiratory Disorder Monitoring
• CERN Zenodo – Adaptable Spam Filter Modelling
• Automated Forensic Ink Recognition via MSP Analysis
• Worm tracking for personality detection
• Decoding Brain States Triggered via Video Game with fMRI Based MVPA Approach
• Machine Learning for biomolecular design and drug development : Predicting the effect of mutation in a protein sequence
• Mandatory Lane Change Prediction
• Modelling of the neutral hydrogen in galaxies with a Fully-Connected Neural Network
• Sleep Stage Classification through Non-Invasive Sensors
• Physics-informed neural networks for a system of discrete masses and springs
• A New 1D CNN Method to detect Sleep Apnea using a single-lead ECG signal
• MultiModal Modular Network (MoMoNet) for predicting COVID diagnosis from mixed image and tabular data
• Machine Learning in Quantitative Finance – Fast Derivatives Pricing and Hedging
• HOPE-Generator
• Intrusion Detection System for Blockchain Transactions
• Modular Decision Support Networks (MoDN) to predict respiratory disease from lung sounds
• Deep Learning-Based Clustering of Images of Damaged Buildings
• Exchange Rate Forecasting with Neural Networks
• Sea ice coverage
• Machine learning in finance: forecasting and trading
• Segmentation of fission yeast microscopy images
• Improving Causality Understanding of Trajectory Prediction Models Using Contrastive Learning
• Predicting Depression From Passive Phone Data – NoCommonPoint -SMSL
• Label-free identification of neurodegenerative aggregates using Deep Learning
• The force applied by a standing wave acoustic levitator on small objects
• On the Limits of Human-centered Commonsense in Large Language Models
• Cell segmentation
• Prognostic models on an international data set of 800’000 patients suspected of COVID-19
• Brain Fingerprinting: Identifying Individuals via learned brain graphs
• Federated learning for semantic segmentation of cracks in images
• Crack Segmentation on FRP CT Images Using Residual Multi-scale Dilated Convolution and Reversed Cross Entropy Loss
• Deep Learning-Based Dysarthric Speech Classification In Adverse Acoustic Environments
• Language Proficiency and Authorship Classification Strength
• ML4Science: Mining Effective Words For Climate Change Communication
• Modelisation of electron diffusion in water for radiation therapy using Generative Adversarial Networks
• Epilepsy Detection
• Doping detection using athlete’s biological passports
• ML4Science – Training an LCD model on the KITTI dataset
• Detection of protein aggregates in neural cells in Parkinson’s disease
• Predicting if two frames are part of the same video
• Adaptive Quantization for 6DoF Pose Estimation
• Costbuster
• Vote Prediction Algorithm – Valais Constituante
• Centriole elimination program in C. elegans
• Firing rate prediction in optic nerve fibers
• Novelty Detection in the Dotcom Bubble
• Do CNNs and Vision Transformers learn visual representations similar to those of the monkey brain?
• Predicting Protein Solubility from Amino Acid Sequences
• Aerospace Bearing
• ML4Science: Cells and Pili Detection on ISCAT images
• Mining Effective Strategies for Climate Change Communication through NLP
• Towards Commonsense Causal Reasoning with Large Language Models
• Bias Slayer: AI4Health – Spotting biased applications
• Fine-tuning a neural language model to address text constraints
• Tracking worms for reading minds by Andromeda
• Budding and Fission Yeast Segmentation
• Combinatorial optimization with Autoregressive Neural Networks
• Predicting the Fatigue Life of Wind Turbine Adhesives from Fracture Surface Images using Machine Learning
• Cell segmentation
• A convolutional neural network segmenting fission yeast microscopy images
• Prediction of fractional flow reserve based on CT scan images and lesion masks of coronary arteries
• Novelty Detection and Revenue Estimation
• Brain Fingerprinting and Task Decoding via Graph Structure Learning
• Wildfire Susceptibility Mapping
• GFP classification project
• Cell segmentation using time-sequence data
• Novenzyme Stability Prediction (Kaggle Competition)
• ML4 Science Physics Informed Neural Networks
• Composer Classifier on MuseScore Sheet Music
• Predicting apartment demand using Machine Learning models
• Predicting Housing Demand
• ML project with the Laboratory of the Physics of Biological Systems
• Estimating the robustness of deep learning fraud detection on the ethereum blockchain against attackers with a leaked model
• Patent Classification into Sustainable Development Goals defined by the United Nations.
• Extending a model to predict poverty through time in African countries
• Characteristics of different transcription factors
• Optimization of Biological Age Prediction from Brain Anatomical Volume Measurements using Subgrouping Models
• Explaining the Valais Constituent Assembly Votes
• Finding relevant gene for centriol assembly and desasembly in C.Elegans

Reproducibility Challenge:

• Reproducibility Study of ”Amortized Tree Generation for Bottom-up Synthesis Planning and Synthesizable Molecular Design”
• Reproducibility Study of ”Scaled-YOLOv4: Scaling Cross Stage Partial Network”
• Grokking – Reproducibility Report
• Reproducibility Study of “Assessing Generalization via Disagreement”
• Reproducibility Study of Behavior Transformers
• Reproducing results for Adaptive Stochastic Variance Reduction for Non-convex Finite-Sum Minimization method
• Numerical influence of ReLU’(0) on backpropagation
  Deep hedging

2021:
ML4Science:

Reproducibility Challenge:

• Re: AdaAttN: Revisit Attention Mechanism in Arbitrary Neural Style Transfer
• Re: Reproducing Empirical Results and Assessing Theoretical Guarantees of PAGE
• Re: Contrastive Learning of Musical Representation
• Re: Invariance Principle Meets Information Bottleneck for Out-of-Distribution Generalization
• Re: Learning by Turning: Neural Architecture Aware Optimisation

2020:
ML4Science:

2019:
ML4Science

• Machine Learning for Invisibility Cloak Architecture
• Predicting PV Array Power Output Using an All-Sky Camera
• Predicting bacteria evolution during cell cycle observations using a Recurrent Neural Network
• Searching for Similar YouTube Channels
• Stress Classification from Biosignals using Neural Networks
• Crystal-Structure Descriptor for Binary Materials Based on Coordination Numbers
• Polymers identification using nanopores
• ML for Science – Generate images of Mesopotamian artifacts
• Machine Learning for Toxicological Testing
• Global gene expression analysis: determine hormone signalling activation in human breast cancer samples
• Yeast Cell Segmentation with U_Net: Effects of Weight Maps and Attention Gates
• Machine Learning for LHC performance optimization
• Product Life Cycle Prediction Using Machine Learning
• Clustering estrogen receptor-positive breast cancer tumors based on hormonal response type
• Identifying the shape of the worms based on the hand-annotated images dataset
• Applause Recognition of Live Concert Recordings for the MetaMedia Center
• Machine Learning for EDS Data Decomposition
• Twitter Astroturfing Detection
• Removing Noise From Microscope Images Without Ground Truth
• U-Net for Yeast Cell Segmentation
• C elegans segmentation
• FCN for neuronal semantic segmentation
• Machine Learning Project 2 – Optimal Option Exit Strategies with Neural Networks
• Bio Product Classifier
• Segmenting yeast cells from microscope images
• Instance Segmentation of Yeast Cells using Mask R-CNN
• A machine learning approach to determine chemical shifts in NMR spectroscopy data
• Detecting rooftop solar PV installations using CNN
• A Physical-Interactive Pac-Man Game for Stroke Rehabilitation
• Logistic Hairdressing
• Prediction Forces for a Flapping Wing system using Linear Regression and Neural Networks Methods
• Exploratory Analysis with MOOC data used for Blended Learning
• Predicting Energy Building Consumption
• Anomaly detection for energy consumption time series
• Segmentation of Cracks on Laboratory Images
• The _place_ (physical and conceptual) of advertising in video game magazines
• Image Classification: Distinguish Murine from Human Cells in PDXs Models
• PLIER for single-cell RNA sequencing data
• Identifying Outer Divertor Legs On Images From Tokamak Experiments
• Gravitational Lens Finding
• VNAV – No GPS, No Problem
• Earthquake Detection from Seismological Data
• Detecting Astroturfing Bots on Twitter
• Feature identification of MANTIS data
• ObsBox Instability Clustering
• Photovoltaic Power Production in Swiss Communes
• Detection of Strong Gravitational Lenses with Convolutional Neural Networks
• Nano Manufacturing with ML and Raman Spectroscopy
• Motion Prediction on Drosophila Using a Seq2seq Model
• Segmentation of spinal cord images
• Avalanche Intelligence
• Pac-Man Error Project
• Building Classification using Google Street-View
• U-Net application for yeast cell segmentation
• Landmark Based Visual Navigation for Drones
• Classification: Crashes and Disruptions in Plasma Experiments
• Automating Route Setting in Climbing via Deep Generative Models
• Leading Edge Suction Prediction of a Dynamically Pitching Airfoil with Trailing Edge Flap
• Aerodynamical parameters estimation from velocity, pressure and noise sensors data using temporal convolutional neural network
• Machine Learning for Science: Diffraction detection in rock structures
• Classifying amino acid modifications using nanopore sequencing data
• Microscopical Image Restoration of C. elegans using Noise2Noise
• ML challenge : Strong Gravitational lensing
• Instance segmentation
• Invisibility cloak
• Pacman collaboration with CHILI lab
• Enchordings – Harmony Embeddings

Reproducibility Challenge:

• Subspace Attack: Exploiting Promising Subspaces for Query-Efficient Black-box Attacks
• Shape and Time Distortion Loss for Training Deep Time Series Forecasting Models
• Unsupervised Object Segmentation by Redrawing
• Face Reconstruction from Voice using Generative Adversarial Networks

2018:
ML4Science

• Quality of Life in Swiss Cities based on OpenStreetMap
• Compressive Sensing MRI using Deep Learning
• Predicting the density of lightning activity from atmospheric and geographic features
• Machine Learning for Science: Quantum Machine Learning
• Solar Panel Recognition and Segmentation on Swiss Map using Convolutional Neural Networks
• ML for crystal structure determination as an alternative to NMR spectroscopy
• Human Behavior Modelling
• Domain-invariant defect detection using deep learning
• Comparing classification techniques for metabolic kinetic models selection
• Spatially-Inferred Graphical Models for fMRI Data Mining
• Crowded enzyme kinetics using simulation and machine learning
• Quality of life in Swiss Cities
• Ultrathin section segmentation
• Correlations between cognitive performance and sensory stimuli in the work environment
• Autism Diagnostic based on Machine Learning
• Human performance modelling according to indoor temperature and light (quantity and colour)
• Automatic Harmonization using Recurrent Neural Networks
• Chord recognition on Beethoven string quartets
• Machine Learning Privacy
• Quality of Life Clustering of Swiss Cities from Insurance and Demographic Data
• Machine learning for air quality measurement and modeling
• The Case for Bagged Neural Networks: Evidence from Outlier Detection using Autoencoder Ensembles
• Chord Prediction with The Annotated Beethoven Corpus
• Predicting Forces on a Flapping Wing Model using Machine Learning
• Brain Tissue Segmentation
• Clustering and Predicting Swiss cities based on Insurance Data
• Predicting the material properties of the arterial wall of a mouse
• Predicting organic carbon with infrared spectra
• 3D Pointclouds Super-resolution for Digital humanities
• Segmentation Of Silicon Wafers For Electron Microscopy Using Mask-RCNN
• Classifying Nanopore Readings with Deep Learning
• Wind Profile Prediction in an Urban Canyon: a Machine Learning Approach
• A Stem Cell Classifier for Single Cell RNA Sequencing Data
• Deep Convolutional Neural Networks for Cell Segmentation in Bright-Field Microscopy Images
• Predicting Aerosol Particles: Sulfate, Nitrate and PM2.5
• Healthy aging: age group prediction from chunking strategies during motor sequence learning
• Multi-lingual text classifier for social media data
• Implementation of an Improved Model for the Prediction of Effective Rate Constants in the Presence of Crowders
• Architecture of Feelings
• Evaluating the quality of videos through machine learning
• Classifying segmentation defects in mutant zebrafish embryo
• Fingerprinting DNS-over-HTTPS traffic
• Statistics of Turkish researchers after the 2016 Coup d’Etat attempt
• Analysis of the dismissal of Turkish researchers after the 2016 Coup d’Etat attempt
• Evaluating the risk of relapse in melanoma

• Improving Generalization and Stability of Generative Adversarial Networks
• Meta-learning with differentiable closed-form solvers
• A Resizable Mini-batch Gradient Descent based on a Multi-Armed Bandit
• Detecting Adversarial Examples via Neural Fingerprinting
• Learning Neural PDE Solvers with Convergence Guarantees
• AutoLoss: Learning Discrete Schedules for Alternate Optimization
• MAE : Mutual Posterior-Divergence Regularization for Variational Autoencoders
• Hyper-Regularization: An Adaptive Choice for the Learning Rate in Gradient Descent