Reaction time prediction¶
Shirazi2024 (HBN)Usage¶
neuralbench eeg reaction_time
Show config.yaml
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
data:
study:
source:
name: Shirazi2024Hbn
filter_task:
name: QueryEvents
query: "task=='task-contrastChangeDetection'"
filter_reaction_time:
name: QueryEvents
query: "(reaction_time >= 0.5) | type == 'Eeg'"
split:
name: PredefinedSplit
test_split_query: "release in ['R5']"
col_name: split
valid_split_by: release
valid_split_ratio: 0.091 # 1/11
valid_random_state: 33
target:
=replace=: true
name: EventField
event_types: Keystroke
event_field: reaction_time
aggregation: trigger
trigger_event_type: Keystroke
start: 0.5
duration: 2.0
summary_columns: [release]
brain_model_output_size: &brain_model_output_size 1
trainer_config.monitor: val/pearsonr
trainer_config.mode: max
loss:
name: MSELoss
metrics: !!python/object/apply:neuralbench.defaults.metrics.get_regression_metric_configs
- *brain_model_output_size
Description¶
This task corresponds to Challenge 1 (“Cross-Task Transfer Learning”) of the EEG Challenge 2025 [Aristimunha2025]. The goal is to predict where, within a 2-s EEG window, participants responded to a visual stimulus with a mouse click.
The official challenge metric is the normalized root-mean-squared error (RMSE divided by the standard deviation of the ground-truth reaction times), reported as normalized_rmse alongside the other regression metrics.
Dataset Notes¶
Shirazi2024 (HBN) contains EEG recordings from 11 cohorts (“releases”) containing different participants. Here, we leave one release out for testing.
The dataset contains different tasks (resting-state, contrast change detection, etc.). Here, we use the contrast change detection (CCD) data as in [Aristimunha2025].