Flagship processed photometry data with behavior labels

<p dir="ltr">Multifiber.pickle contains a dictionary featuring every recorded session, from training period to hard fight.<br>Each key has the following structure:</p><p dir="ltr">animalID_day_condition_session,<br>Where animalID is the animal identifier, day ranges from d1 to d9, condition identifies the behavior group (balbc or mCD1 for EXP, obs or OBSmCD1 for OBS, toy or toyCD1 for NON), and session identifies the session recorded within-day (ranges from t1-t3 as there are 3 sessions in each day).</p><p dir="ltr">Each key contains a dataframe table, wherein each column denotes a neural population recorded (VMHvl, PMv, AH, POA, BNST, MeA, PA, lPAG, vLS, PrL, LHb, NAc DA), each identified by cell type ( (E) denoting excitatory, (I) denoting inhibitory, (DA) denoting dopaminergic for NAc).

Additionally, columns 'unsupervised labels' includes categories [1-13] denoting behavior clusters frame-by-frame obtained from unsupervised behavior classification (UMAP -> filtering -> watershed from SLEAP-derived features) and 'attack labels' contains a binary set [0,1] denoting attack behaviors frame-by-frame obtained from supervised classification (MLP trained on SLEAP-derived features to predict hand-annotated attacks).<br>Note that columns with nans were assigned as such deliberately due to histological exclusion criteria.</p>