Supplementary Figure 5: Single-trial factor trajectories from the multisession stitched LFADS model for a representative session.
From: Inferring single-trial neural population dynamics using sequential auto-encoders
LFADS factors are projected into a space consisting of the condition-independent signal (CIS) and the first rotational plane identified using jPCA; this is the same projection used in Fig. 4d in the main text. Black dots depict the time of go cue presentation, and each trajectory proceeds for 510 ms. Colors indicate reach direction as in Fig. 4. Thin traces are trajectories from individual trials; thick traces of the same color are the condition-averaged trajectory for each reach direction.
