Difference between revisions of "Course syllabus: Human-computer interfaces"

From Research management course
Jump to: navigation, search
(Created page with "Multi-model selection and heterogenous data analysis with application to Human-computer interfaces A significant part of the data measurements relate to the physical world: v...")
 
 
Line 1: Line 1:
Multi-model selection and heterogenous data analysis
+
{{#seo:
with application to Human-computer interfaces
+
|title=Human-computer interfaces - course syllabus
 +
|titlemode=replace
 +
|keywords=Human-computer interfaces
 +
|description=Multi-model selection and heterogenous data analysis with application to Human-computer interfaces
 +
}}
 +
Multi-model selection and heterogenous data analysis with application to Human-computer interfaces
  
 
A significant part of the data measurements relate to the physical world: video, audio, weather measurements, human fMRI and iEEG, and even metaverse data. Revealing dependencies between these heterogeneous measurements is the most acute problem in Machine learning. We have to select a model that maps the design state of a dynamic system to its target state, given its behavioral history and environment. In Deep learning, it is a sequence-to-sequence problem. Since a dynamic system observes time, we must use Geometric deep-learning statements.  
 
A significant part of the data measurements relate to the physical world: video, audio, weather measurements, human fMRI and iEEG, and even metaverse data. Revealing dependencies between these heterogeneous measurements is the most acute problem in Machine learning. We have to select a model that maps the design state of a dynamic system to its target state, given its behavioral history and environment. In Deep learning, it is a sequence-to-sequence problem. Since a dynamic system observes time, we must use Geometric deep-learning statements.  

Latest revision as of 23:51, 13 February 2024

Multi-model selection and heterogenous data analysis with application to Human-computer interfaces

A significant part of the data measurements relate to the physical world: video, audio, weather measurements, human fMRI and iEEG, and even metaverse data. Revealing dependencies between these heterogeneous measurements is the most acute problem in Machine learning. We have to select a model that maps the design state of a dynamic system to its target state, given its behavioral history and environment. In Deep learning, it is a sequence-to-sequence problem. Since a dynamic system observes time, we must use Geometric deep-learning statements.

This course aims to create an impactful software library for sequential multi-modeling. The main stages of this course are researching state-of-the-art solutions, choosing appropriate data and applications, gathering and testing basic libraries, assembling the system, and disseminating the results to collect responses from the community of researchers. Since this is a group project, planning the members’ roles is important. It must deliver the results within time and resources, maximizing the outcome.

The course plan contains practical steps:

  1. Introduction to the problem, discussing the theoretical statements
  2. Discussing the state-of-art libraries and choosing datasets, getting the role responsibilities
  3. Composing the library architecture, writing and discussing the interfaces
  4. Testing parts of the library given fixed data on an extreme-programming basis
  5. Presentation of the personal results for the system parts
  6. Align the interfaces and discuss the trivial and the minimum valuable systems
  7. Assemble the system so that every group member tests various subsets of models
  8. Presentation of the system tests, making a comparison table with the results of the test
  9. Writing the tutorial on the system with concepts, theory, and examples of usage
  10. Selecting the target reviewers, connecting them, and getting responses
  11. Gathering and processing responses, preparing the dissemination materials
  12. Publication of the library, its video presentation with slides, introductory materials, and tutorial

Since this course gives four credits, the homework should take the lowest of four hours of qualified work weekly. The contribution of each group member counts in the designated table along with the peer review.

Scoring: each student gets weekly homework. The results go to the repository and in the peer review table. The summary is the student’s weekly checkmarks and peer-reviewed contribution.

The students obtain the following knowledge and skills. Knowledge of multi-model selection techniques, heterogeneous data analysis, geometric deep learning, and physics-informed learning. Skills in project planning, estimating impact and consequences, system design and testing, group role planning, peer review, and technical communication.

Informal rules of communications

  • Style
    • No coding, just assembling
    • Orientation to the user and reader
  • Rules of system creation:
    • Each player could be omitted without system degradation
    • The trivial system exists in the beginning, so the first who contributes gets the prize
  • Rules of scoring:
    • It is better to have automatic scoring
    • Table score for minimum and peer review or the final score