Dr. Omar Costilla Reyes studies interpretable machine learning for digital mental health, neural and sensor data, symbolic modeling, and scientific discovery.
These research programs connect sensor and neural data, digital health studies, and methods for building models and programs that researchers can inspect.
Studies of smartphone-delivered interventions, treatment response, mobile sensing, and behavioral patterns.
Explore this areaResearch on program synthesis, reusable concepts, and equation discovery for inspectable models.
Explore this areaWork with floor sensors, gait and footstep signals, and pattern recognition across sensing applications.
Explore this areaResearch on electrophysiology, neural rhythms, cognitive tasks, and computational analysis of brain signals.
Explore this areaResearch and engineering work on interpretable systems, health data, and human review.
Explore this areaResearch on how predictive models in digital mental-health studies can be evaluated, explained, and interpreted by domain experts.
How can researchers examine what a model learned from digital mental-health data?
Explore this area
Research on using symbolic programs and equation-search methods to turn experimental data into candidate models that researchers can inspect and test.
How can experimental data support candidate scientific models and equations?
Explore this area
Research on searching numerical data for equations that can be inspected, compared, and tested against the original scientific problem.
How can equation search find useful, testable structure in numerical data?
Explore this area
This catalog connects current research with earlier engineering, sensing, robotics, and community projects. Each entry links to a publication or project source when one is available.
Equ Healthcare develops individual-focused digital health services that bring wearable, laboratory, lifestyle, and behavioral information into a clearer personal view. The work emphasizes privacy-conscious data use and interpretable outputs that help people understand the information behind health-related insights.
This project applies interpretable machine learning to smartphone-delivered mental-health interventions, including work on depression improvement and treatment response. Related studies examine how mobile sensing, clinical measures, and anxiety comorbidity relate to outcomes; the models are research analyses rather than diagnostic tools.
This MIT research program combines machine learning with symbolic programs, reusable concepts, and scientific domain knowledge. The aim is to build computational methods that can discover useful mathematical structure while leaving models and reasoning steps available for inspection by researchers.
InceptionSR reuses partial equations as features within a recursive symbolic-regression search. The AAAI 2025 study evaluates this approach on equation-discovery tasks, including a planetary-instability problem, and examines how reusing learned structure affects model fit and the readability of candidate equations.
This NeurIPS 2024 work studies how a large language model can help discover and reuse abstractions found in high-performing equations. The method combines evolutionary search with concept-guided steps and is evaluated on Feynman equations and synthetic tasks; reported results apply to those benchmarks.
This study compares grammatical-evolution methods for estimating total body-fat percentage from anthropometric measurements. It focuses on the trade-off between predictive performance and equations that can be read and examined; the resulting estimates are research models, not individual clinical assessments.
This collaborative research project examines how complementary measurements can support preclinical and early Alzheimer’s research. Its planned phenotyping combines amyloid and tau positron-emission tomography (PET), structural magnetic resonance imaging (MRI), retinal imaging, and biomarkers for experimental medicine and clinical-trial research.
Research in digital phenotyping uses patterns from mobile devices and repeated self-reports to study behavior over time. This work applies interpretable modeling and network analysis to schizophrenia-related measures, with findings bounded by the study data and not intended as a clinical diagnosis.
This research uses floor-sensor recordings and deep-learning methods to analyze the spatial and temporal structure of footsteps. It connects gait analysis with healthcare and biometric-security research while treating reported recognition performance as specific to each study’s dataset and evaluation design.
Micro Atlas investigated drone-based sensing for signs of crop disease, drought, pests, and nutrient shortages. The project brought imaging and sensor analysis to agricultural monitoring questions, with the aim of identifying field conditions that warrant closer inspection.
This robotics project implemented and optimized an RGB-D (red, green, blue, and depth) simultaneous localization and mapping method for an indoor mobile robot. A Kinect camera supplied real-time video for mapping the environment and estimating the robot’s location within it.
Developed as an Android and iOS proof of concept for downtown Dallas, Walk This Way mapped routes between two places according to user-selected criteria, including safety, shade during the day, light at night, and architecture.
As a graduate assistant in the Research Experience for Teachers (RET) program, Dr. Costilla Reyes supported high-school teachers developing an engineering research project with cooperative line-following robots. The program connected classroom educators with hands-on science, technology, engineering, and mathematics research.
This project developed a rover-navigation approach using stereo vision, laser-range measurements, and infrared sensors. A MATLAB algorithm used three-dimensional terrain maps and sensor readings to guide movement around detected objects; the design was evaluated with a Pekee II robot and its available robotics APIs.
Dr. Costilla Reyes co-organized a summit at the MIT Media Lab that brought Latin American policymakers and AI researchers together to discuss artificial-intelligence adoption in the region. The January 2020 program addressed applications across areas including health, education, climate, and economic development.