This dataset contains sorted unit spiking times and behavioral data from a macaque performing a reaching task with perturbations. In the experimental task, the subject performed delayed center-out reaches using a manipulandum to control a cursor. On a portion of the trials, the manipulandum applied a bump during the center hold prior to the reach. Neural activity was recorded from an electrode array implanted in somatosensory area 2. Hand position, cursor position, force applied to the manipulandum, length and velocity of various arm muscles, and angle and velocity of various arm joints were all recorded during the experiment. Provided as part of the Neural Latents Benchmark: https://neurallatents.github.io.

This dataset contains sorted unit spiking times and behavioral data from a macaque performing a delayed reaching task. The experimental task was a center-out reaching task with obstructing barriers forming a maze, resulting in a variety of straight and curved reaches. Neural activity was recorded from electrode arrays implanted in the motor cortex (M1) and dorsal premotor cortex (PMd). Cursor position, hand position, and eye position were also recorded during the experiment, and hand velocity was calculated offline from hand position. Provided as part of the Neural Latents Benchmark: https://neurallatents.github.io.

The upper and lower block display model performance (mean ± std.error) when reconstructing face images and natural images, respectively, in terms of six metrics of perceptual cosine similarity using the five MaxPool layer outputs of VGG16 for face recognition (face images) / object recognition (natural images) and latent cosine similarity between w-latents of stimuli and their reconstructions. The rows display decoding performance when using the recordings from all recording sites (i.e., V1, V4 and IT together) or the recordings within a specific brain area.