Notes

Methods, concepts, and context for a closer reading.

Focused explanations, figures, and equations that support the projects and writing. Read them on their own or open them alongside an article when you want more detail.

A BP prediction pilot versus clinical validation AI

How to interpret guidance reference values without claiming a validation pass.
BCG: mechanical activity measured through the bed AI

Film and load-cell channels are two ways of measuring the heartbeat mechanically.
Blood pressure: systolic and diastolic AI

What the two pressure targets mean and how they are measured in this project.
Channels, responses and feature maps AI

Reading a spatial channel as one learned computation applied across image locations.
Choosing the rank of a dynamical model AI

How predictive dimension, estimation stability and held-out performance inform model rank.
Clustering regions and detecting graph communities AI

Why grouping similar profiles and partitioning graph connections can produce different network organizations.
Comparing subjects in a shared latent basis AI

Why common ROI locations do not guarantee comparable latent dimensions, and what a shared basis makes comparable.
Continuous dynamics and discrete states AI

Why signed predictive relationships between continuous components are different from probabilities of switching states.
Contribution magnitude and variance explained AI

Why the size of a component contribution is not an additive percentage of the recorded signal.
ECG: an electrical timing reference AI

Why ECG supplies cycle boundaries and timing information for pressure estimation.
From a representation to class scores AI

How spatial averaging and a linear classifier connect the image maps to predictions.
From fMRI signals to network graphs AI

How measurement units, edge definitions and graph summaries shape the meaning of a brain network.
From scientific abstracts to a topic map AI

Why the atlas uses one reduced representation for clustering and another for display.
From waveform landmarks to model features AI

Timing, magnitude and shape measurements, and the 42-feature selection.
Hidden units and hidden states AI

How one numerical response relates to a complete representation and its history.
How Ward clustering becomes a topic hierarchy AI

The difference between the fitted binary dendrogram and the topic branches exposed in the viewer.
How a GRU carries information through time AI

How a recurrent update turns nine sensor values into a changing 128-value state.
How rising and falling topics are screened AI

Combining a year-score test, endpoint effect size and rank consistency to identify directional topic histories.
How short lived topic peaks are detected AI

Comparing interior windows with both neighboring periods to distinguish local peaks from sustained rises.
How topic branches receive readable names AI

Using keywords and sampled papers to name a branch without treating the name as evidence for its contents.
Input scaling and batch normalization AI

Distinguishing fixed sensor scaling from normalization layers inside an image network.
MAE, bias and error spread AI

How the reported error measures differ, and what equal weighting of participants means.
PPG: the optical pulse waveform AI

What finger photoplethysmography adds to the electrical and mechanical signals.
Parcels and functional systems describe different scales AI

How regional measurement units, distributed reference systems and learned network patterns relate.
Personal calibration and later-beat evaluation AI

What the models learn early, what is held out and why five-beat windows do not overlap training.
Personal regression versus cohort sharing AI

How partial pooling borrows information while preserving individual calibration.
Reading a map of learned representations AI

How to use UMAP proximity without mistaking a display for a classifier or explanation.
Reading pressure scatterplots and agreement plots AI

What the interactive axes and agreement lines show.
Reading topic timelines in a balanced sample AI

What annual counts and peak normalization can show when roughly equal numbers of papers are sampled from each year.
Recordings and reference blood pressure AI

The analyzed participants, signal preparation and reference-quality rules.
Registration, resampling, and anatomical coverage AI

Three distinct questions behind an ROI time series: anatomical alignment, image sampling, and available tissue.
Ridge regression for a personally calibrated pressure model AI

What regularization does, why feature scaling matters and how the personal model is fitted.
Scores, softmax and confidence AI

Why a high displayed probability is not the same as demonstrated reliability.
Second moments, covariance, and connectivity AI

Why a matrix of signal products can retain mean structure that is absent from covariance and correlation.
Separating waveform magnitude, shape and timing variability AI

How normalization, alignment and empirical bands change a waveform comparison.
Spatial resemblance and concentration measure different things AI

Why a strong spatial correlation need not mean that most of a pattern lies inside a reference system.
Spectral learning of predictive dynamics AI

How past–future relationships become a compact dynamical representation and spatial network patterns.
The four-feature source-study comparison AI

What the approximate reproduction preserves and how its evaluation differs.
Training, weights and checkpoints AI

Separating model updates from responses to one fixed input.
Visual systems beyond a single visual network AI

How retinotopic maps, processing streams and category-selective regions relate to broad visual reference labels.
What ReLU does to a response AI

Why some image-network responses are zero and why that differs from a GRU state.
What block and resting state designs reveal about dynamics AI

How an experimental clock changes the interpretation of network dynamics, including the limits of fixation baselines.
What different connectivity measures ask AI

A practical distinction between correlation, conditional association, frequency-specific coupling and directed models.
What fine-tuning changes AI

Distinguishing pretrained features, a new classifier and full-network adaptation.
What holding out participants tests AI

Why activity windows must be separated by person when evaluating new-person performance.
What might resting state activity be doing AI

Evidence and complementary interpretations of intrinsic organization, self-generated thought and postexperience processing.
What the atlas contours represent AI

How smoothed map coordinates become navigable shapes and why contour size is not topic size.
Why ResNet uses residual blocks AI

How a learned transformation and a shortcut combine within the image network.
Why screening many topics needs a correction AI

Interpreting adjusted p-values when related topic branches and multiple temporal patterns are examined together.
No matching items