Figure 1.
Models and experimental data are collected from different labs and aligned to a common format that allows analyses using DREAM tools. Lighter solid boxes indicate a standard model format, darker solid boxes indicate a standard data format (for experimental data and model predictions), and striped boxes indicate DREAM tools. Because the model and data formatting is consistent throughout DREAM, tools can work for any model, experimental dataset, or model prediction.
Figure 2.
Publications from which data has been added to DREAM. The DREAM ID is a unique identifier created by combining the last name of the first author and the year of publication.
Figure 3.
Hierarchical structure of experimental data in DREAM.
Red dotted lines indicate that more fields could be present, but were omitted for space. Blue lines indicate meta-information; green lines indicate recorded experimental data. The contents of ‘Special’ are not standardized and can contain any arbitrary data. The ‘Subject’ field may also have a ‘Special’ field, depending on the experiment.
Figure 4.
Preferred velocity of two neurons.
The x–y velocity of the hand when a spike occurred. This shows two different neurons from two different monkeys performing the same center-out task.
Figure 5.
Two DREAM tools used to visualize experimental data and model predictions.
Hand trajectories and velocity profiles from subject 13 of the Mattar_2007_e3 dataset, and the predictions of the same dataset using the Berniker_2011_m1 model. The force field region has been overlaid on the plot for clarity. For the velocity profiles, five reaches are displayed. One reach is the first force field the subject experienced (315°). We also show the velocity profile for the first and last catch trial the subject experienced (also at 315°), the last force field trial (315°) and the first null test trial, which involved reaching to the 225° target after learning the force field at the 315° target.
Figure 6.
Mean Squared Error Averaged Across Subjects.
The results of four separate models run against all experiments in DREAM. The Mean Squared Error (+/− STD) averaged across subjects for each experiment is reported.