Middle school
Vocabulary
Every new word in this book, and where you meet it.
- Alpha
- A rhythm of about 8 to 13 waves a second, strongest at the back of the head with the eyes closed.
- Lab 1, Alpha Flower
- Artifact
- A signal in a recording that is not from the brain, like a blink or a clenched jaw.
- Chapter 3, Signal and noise
- Brain-computer interface
- A system that turns a signal from the brain into a command, with no muscles in between.
- Chapter 1, The brain's electricity
- Consent
- Agreeing after understanding what will happen, without pressure, with the freedom to change your mind.
- Chapter 4, Your brain data
- Custom AI head
- The small model you teach on the Mind-Controlled Light Switch. It gives each reading a score, near 0 for Off and near 1 for On.
- Lab 4, Mind-Controlled Light Switch
- EEG
- Electroencephalography: the brain's electrical activity, recorded from electrodes on the scalp.
- Chapter 1, The brain's electricity · Lab 1, Alpha Flower
- Electrode
- A small contact that touches the skin and picks up electrical signals.
- Chapter 1, The brain's electricity · Lab 1, Alpha Flower
- Engineering design process
- Define the problem, imagine ideas, plan one, create it, test it, improve it, and test again.
- Chapter 8, Designing a BCI
- LED
- A small light that lights when current flows through it, one way only.
- Chapter 7, Microcontrollers and circuits
- Microcontroller
- A small computer on one chip, built to do one job.
- Chapter 7, Microcontrollers and circuits
- Model
- A set of rules a computer learns from examples, so it can decide about new ones.
- Chapter 5, Models that learn
- Neuron
- A brain cell that sends signals to other cells with pulses of electricity.
- Chapter 1, The brain's electricity · Lab 1, Alpha Flower
- Pin
- A connection on the board that a program can switch on, to 3.3 volts, or off.
- Chapter 7, Microcontrollers and circuits
- Training and testing
- Training fits the rules to examples whose answers are known. Testing measures them on examples held back.
- Chapter 5, Models that learn
- Visual cortex
- The part of the brain at the back of the head that handles what you see.
- Lab 2, BCI Flight Simulator
A
B
C
D
E
F
H
L
M
N
P
R
T
V
High school
Vocabulary
Every new term in this book, and where you meet it.
- 10–20 system
- The standard map of electrode positions on the scalp, set by fractions of the distances between the nasion, the inion and the points in front of the ears.
- Lab 1, Alpha Flower
- Action potential
- The brief electrical pulse that travels along a neuron.
- Chapter 1, The brain's electricity
- Advertising and pairing
- Broadcasting what a device is, and setting up the keys that encrypt a link.
- Chapter 8, Microcontrollers and circuits
- Alpha
- A rhythm of about 8 to 13 waves a second, strongest at the back of the head with the eyes closed.
- Lab 1, Alpha Flower
- Artifact
- A signal in a recording that is not from the brain, like a blink or a clenched jaw.
- Lab 1, Alpha Flower · Chapter 3, Signal and noise
- Baseline
- Something to compare against: a simple model fitted the same way, or a measurement taken before anything changes.
- Lab 2, BCI Flight Simulator · Lab 3, Age Detector
- Binomial distribution
- How often each number of successes comes up, in a set number of independent tries with the same chance each time.
- Chapter 6, Chance and fair tests
- Consent
- Agreeing after understanding what will happen, without pressure, with the freedom to change your mind.
- Chapter 4, Your brain data
- Custom AI head
- The small model you teach on the Mind-Controlled Light Switch. It gives each reading a score, near 0 for Off and near 1 for On.
- Lab 4, Mind-Controlled Light Switch
- EEG
- Electroencephalography: the brain's electrical activity, recorded from electrodes on the scalp.
- Chapter 1, The brain's electricity
- Electrode
- A small contact that touches the skin and picks up electrical signals.
- Chapter 1, The brain's electricity
- Encoder
- The AI model on the IntoMind One's chip that sums up four seconds of EEG.
- Chapter 7, Encoders and heads
- Engineering design process
- Define the problem, imagine ideas, plan one, create it, test it, improve it, and test again.
- Chapter 9, Designing a BCI
- Error
- How far a guess is from the truth. The mean error is the average size of the errors, ignoring direction.
- Lab 3, Age Detector
- False positive and false negative
- A false positive says yes when the truth is no. A false negative says no when the truth is yes.
- Chapter 9, Designing a BCI
- Frequency tagging
- Giving each choice its own flicker rate, so the rate in the EEG says which one has your attention.
- Lab 2, BCI Flight Simulator
- Head
- A small model that turns an embedding into an answer, such as an age or a score.
- Chapter 7, Encoders and heads
- Independent
- Not affecting each other. EEG readings close in time are not.
- Chapter 6, Chance and fair tests
- Microcontroller
- A small computer on one chip, built to do one job.
- Chapter 8, Microcontrollers and circuits
- MicroPython
- Python made small enough to run on a microcontroller.
- Chapter 8, Microcontrollers and circuits
- Model
- A set of rules a computer learns from examples, so it can decide about new ones.
- Chapter 5, Models that learn
- Motor cortex
- The strip of the brain above and in front of the ears that controls movement. Each side controls the opposite hand.
- Lab 4, Mind-Controlled Light Switch
- Mu rhythm
- A rhythm of about 8 to 13 waves a second over the motor cortex, which drops when the opposite hand moves or imagines moving.
- Lab 4, Mind-Controlled Light Switch
- Neuron
- A brain cell that sends signals to other cells with pulses of electricity.
- Chapter 1, The brain's electricity
- Neurorights
- Proposed rights to mental privacy, mental identity and freedom of thought.
- Chapter 4, Your brain data
- Occipital
- At the back of the brain, where vision is processed. The positions over it are O1, O2 and Oz.
- Lab 1, Alpha Flower
- Overfitting
- Fitting the noise in the training examples, so the rule does worse on new ones.
- Chapter 5, Models that learn
- Permutation test
- Relabeling the data in every way that keeps its structure, to see what scores chance alone produces.
- Chapter 6, Chance and fair tests
- Pin
- A connection on the board that a program can switch on, to 3.3 volts, or off.
- Chapter 8, Microcontrollers and circuits
- Postsynaptic potential
- The longer change in a neuron's voltage when another neuron signals to it.
- Chapter 1, The brain's electricity
- p-value
- How often a result at least this strong would happen if only chance were at work.
- Chapter 6, Chance and fair tests
- Sample rate
- How many measurements of a signal a device makes each second.
- Chapter 2, Waves and rhythms
- SSVEP
- Steady-state visually evoked potential: the brain's response at the rate of a light that flickers steadily.
- Lab 2, BCI Flight Simulator
- Training and testing
- Training fits the rules to examples whose answers are known. Testing measures them on examples held back.
- Chapter 5, Models that learn
- Visual cortex
- The part of the brain at the back of the head that handles what you see.
- Lab 2, BCI Flight Simulator
0 to 9
A
B
C
E
F
H
I
L
M
N
O
P
R
S
T
V
College
Vocabulary
Every term this book defines, and where it is taught.
- 10–20 system
- The standard map of electrode positions, set by fractions of the distances between the nasion, the inion and the points in front of the ears.
- Lab 1, Alpha Flower
- Action potential
- The brief electrical pulse that travels along a neuron's axon, lasting about a millisecond.
- Chapter 1, The brain's electricity
- Advertising and pairing
- Broadcasting what a device is, and setting up the keys that encrypt a link.
- Chapter 9, Microcontrollers and circuits
- Aliasing
- A frequency above half the sample rate appearing as a lower one, because the samples can't tell them apart.
- Chapter 2, Waves and rhythms
- Alpha
- A rhythm of about 8 to 13 hertz, largest over the back of the head in a person awake with the eyes closed.
- Lab 1, Alpha Flower
- Artifact
- A signal in a recording that is not from the brain, such as a blink, a clenched jaw or a moving electrode.
- Chapter 3, Signal and noise
- Baseline
- Something to compare against: a simple model fitted the same way, or a measurement taken before anything changes.
- Chapter 8, Evaluating models
- Bias and variance
- Error from a rule too simple to fit the truth, and error from a rule too sensitive to the particular examples it was fitted on.
- Chapter 5, Models that learn
- Bootstrap interval
- An interval from resampling the units that vary, here studies, many times.
- Chapter 8, Evaluating models
- Characteristic
- A value a Bluetooth device offers, to read, write, or subscribe to.
- Chapter 9, Microcontrollers and circuits
- Consent
- Agreement given freely, after understanding what will happen, and open to withdrawal at any time.
- Chapter 4, Your brain data
- Contact
- How well an electrode touches the skin: the better the contact, the less noise it adds.
- Chapter 3, Signal and noise
- Current dipole
- A current source and sink a short distance apart: the model of a patch of cortex whose aligned neurons are active together.
- Chapter 1, The brain's electricity
- Custom AI head
- The small linear model you teach on the Mind-Controlled Light Switch. It gives each reading a score, near 0 for Off and near 1 for On.
- Lab 5, Mind-Controlled Light Switch
- Data minimization
- Collecting and keeping only what a purpose needs, and processing it where it is produced when possible.
- Chapter 4, Your brain data
- Distillation
- Training a small model to reproduce a larger one's outputs.
- Chapter 7, Encoders and heads
- Encoder
- The AI model on the IntoMind One's chip that sums up four seconds of EEG.
- Chapter 7, Encoders and heads
- Ethics board
- A committee that reviews research with people before it starts, such as an institutional review board.
- Chapter 4, Your brain data
- Head
- A small model that turns an embedding into an answer, such as an age or a score.
- Chapter 7, Encoders and heads
- Independent
- Not affecting each other. EEG readings close in time are not.
- Chapter 6, Chance and fair tests
- Leakage
- Information from the test data reaching the training, in any way.
- Chapter 6, Chance and fair tests
- Leakage and window
- Power spreading from one frequency into others because a signal is cut off, and the taper applied to reduce it.
- Chapter 2, Waves and rhythms
- Mean absolute error
- The average size of the errors, ignoring their direction.
- Chapter 8, Evaluating models
- Microcontroller
- A small computer on one chip, built to do one job.
- Chapter 9, Microcontrollers and circuits
- Model
- A rule with parameters fitted to examples, so it can make a decision about new ones.
- Chapter 5, Models that learn
- Model card
- A short account of a model: what it is for, how it was evaluated, and where it fails.
- Chapter 8, Evaluating models
- Motor cortex
- The strip of cortex in front of the central sulcus that controls movement. Each side controls the opposite side of the body.
- Lab 5, Mind-Controlled Light Switch
- Multiple comparisons
- Many tests at once, each with its own chance of a false positive.
- Chapter 6, Chance and fair tests
- Mu rhythm
- A rhythm of about 8 to 13 hertz over the sensorimotor cortex, which falls when the opposite hand moves or imagines moving.
- Lab 5, Mind-Controlled Light Switch
- Neural data
- Data recorded from the brain or nervous system, such as EEG, and what is derived from it.
- Chapter 4, Your brain data
- Neurorights
- Proposed human rights to mental privacy, mental identity and freedom of thought.
- Chapter 4, Your brain data
- Overfitting
- Fitting the noise in the training examples, so the model does worse on new ones.
- Chapter 5, Models that learn
- Permutation test
- Relabeling the data in every way that keeps its structure, to see what scores chance alone produces.
- Chapter 6, Chance and fair tests
- Pin
- A connection on the board that a program can switch on, to 3.3 volts, or off.
- Chapter 9, Microcontrollers and circuits
- Positive predictive value
- Of the cases a test calls positive, the fraction that are.
- Chapter 8, Evaluating models
- Postsynaptic potential
- The slower change in a neuron's membrane potential when another neuron signals to it, lasting tens of milliseconds.
- Chapter 1, The brain's electricity
- Preregistration
- Recording the plan for a study before collecting its data.
- Chapter 11, Research methods
- p-value
- How often a result at least this strong would happen if only chance were at work.
- Chapter 6, Chance and fair tests
- Quantization
- Rounding a model's weights to small integers so the device's chip can run them.
- Chapter 7, Encoders and heads
- Reference and montage
- The electrode or average each channel is measured against, and the set of differences chosen to look at.
- Chapter 1, The brain's electricity
- Reproducible
- Able to be repeated by someone else, from the same data and code, to the same result.
- Chapter 11, Research methods
- Requirement
- A condition a design must meet, stated so a test can check it.
- Chapter 10, Designing a BCI
- Ridge regression
- Least squares with a penalty on large weights, so a model with many inputs and few examples does not fit the noise.
- Chapter 5, Models that learn
- Signal-to-noise ratio
- The power of what you want to measure, divided by the power of everything else in the same band.
- Chapter 3, Signal and noise
- SSVEP
- Steady-state visually evoked potential: the brain's response at the rate of a light that flickers steadily, and at its harmonics.
- Lab 2, BCI Flight Simulator
- Training and testing
- Fitting a model on examples whose answers are known, and measuring it on others held back.
- Chapter 5, Models that learn
- User-centered design
- Designing with the people who will use a system, from their needs to the evaluation.
- Chapter 10, Designing a BCI
- Volume conduction
- Currents spreading through the tissues of the head, so that every electrode records a weighted sum of many sources.
- Chapter 1, The brain's electricity
- Welch's method
- Estimating a spectrum by averaging the spectra of overlapping, windowed segments.
- Chapter 2, Waves and rhythms
0 to 9
A
B
C
D
E
F
H
I
L
M
N
O
P
Q
R
S
T
U
V
W
Graduate
Vocabulary
Every term this book defines, and where it is taught.
- Aperiodic component
- The part of a power spectrum that falls smoothly with frequency, modeled as an offset, a knee and an exponent, with the rhythms as peaks above it.
- Chapter 2, Waves and rhythms
- AUC
- The area under the ROC curve: the probability that a random member of one group outscores a random member of the other.
- Chapter 8, Evaluating models
- Canonical correlation analysis
- Finding the combinations of two sets of signals, here the channels and sinusoids at a flicker rate, that correlate most, and scoring the rate by that correlation.
- Lab 2, BCI Flight Simulator
- Current dipole
- A current source and sink a short distance apart, with a moment in ampere-meters: the model of a patch of aligned neurons active together.
- Chapter 1, The brain's electricity
- Distillation
- Training a small model to reproduce a larger model's outputs, here its representations of EEG, with no labels needed.
- Chapter 7, Encoders and heads
- Effective degrees of freedom
- The sum over a ridge fit's directions of each eigenvalue over itself plus λ: how many weights the fit effectively spends.
- Chapter 5, Models that learn
- Embedding
- The encoder's description of a four-second window: the mean of its tokens, 76 numbers.
- Chapter 7, Encoders and heads
- Encoder
- The neural network on the device that describes each four-second window of EEG with tokens and a 76-number embedding.
- Chapter 7, Encoders and heads
- Equivalence test
- A test that a difference lies inside a margin set in advance, by two one-sided tests: the way to show an effect is small, which a nonsignificant test can't.
- Lab 3, Age Detector
- Estimand
- Exactly what a study estimates: the quantity, the contrast, the summary and the population, fixed before the data.
- Chapter 11, Research methods
- Head
- A linear map from an embedding to up to 32 outputs, stored as signed-byte weights with a scale and a bias, and run on the device.
- Chapter 7, Encoders and heads
- Individual alpha frequency
- The frequency of one person's alpha rhythm, by the peak or the center of gravity of eyes-closed power in a search window.
- Chapter 2, Waves and rhythms
- Information transfer rate
- Bits a minute from the number of choices, the accuracy and the time per choice, assuming equally likely choices and evenly spread errors.
- Chapter 10, Designing a BCI
- Institutional review board
- A committee that reviews research with people before it begins, and decides whether its risks, consent and data handling are acceptable.
- Chapter 4, Your brain data
- Lead field
- The matrix that maps source dipole moments to the potentials at the electrodes, computed from a model of the head's geometry and conductivities.
- Chapter 1, The brain's electricity
- Leakage
- Information about the test data reaching a model while it is fitted or chosen, which makes its test optimistic.
- Chapter 6, Chance and fair tests
- Minimum-norm estimate
- The source pattern with the smallest total size that explains the data, with a regularization parameter that trades fit for size.
- Chapter 1, The brain's electricity
- Multitaper estimate
- The average of a segment's spectra under several orthogonal Slepian tapers, each concentrated in a band of chosen width.
- Chapter 2, Waves and rhythms
- Permutation test
- A test that compares a statistic with its values under every relabeling that the design makes equally likely when nothing is there.
- Chapter 6, Chance and fair tests
- Power spectral density
- The distribution of a signal's variance over frequency, in squared units per hertz.
- Chapter 2, Waves and rhythms
- Preregistration
- A dated, public record of a study's hypotheses, design and analysis, made before the data exist, which separates the tests it planned from what it found by exploring.
- Chapter 11, Research methods
- Requirement
- A condition a design must meet, with a metric, a threshold, the conditions and the users, stated so a test can check it.
- Chapter 10, Designing a BCI
- Ridge regression
- Least squares with a penalty λ on the weights' squared size, which shrinks each direction of the data by its eigenvalue over its eigenvalue plus λ.
- Chapter 5, Models that learn
- Welch's method
- Estimating a spectrum as the average of the periodograms of overlapping, windowed segments, trading resolution for variance.
- Chapter 2, Waves and rhythms