Sensing motion, the review.
Sensing motion in eleven concepts, from the accelerometer to forecasting, and six questions that bring the lessons together. If an answer doesn't add up, under the explanation you'll find the chapter to read again.
The key concepts.
- Accelerometer
The sensor that measures the pushes on the phone in three directions, x, y and z. When the phone is still it measures gravity, about 9.8 meters per second squared, pointing down.
- Gyroscope
The sensor that measures how fast the phone turns around its three axes, in degrees per second. Together with the accelerometer it gives the six axes of motion.
- Window
A piece of signal, for example two seconds long, that the model looks at all at once. The window slides forward in regular steps and the model answers at each step.
- Features
A few numbers computed from a window that sum it up, like the average of each axis, how much it swings or how many peaks there are. They make very small models possible.
- Gesture and activity
A gesture has a start and an end, like shaking the phone. An activity goes on over time, like walking. This changes how long the window has to be.
- Anomaly
A piece of data that doesn't look like the ones seen during training. An anomaly model learns what normal looks like and flags whatever strays from it, without needing any examples of faults.
- Autoencoder
A network that squeezes the input into a few numbers and tries to rebuild it. Trained on normal data, it copies strange data badly, and the error of the copy becomes the anomaly score.
- Drift
When the data changes slowly over time, for example a machine that gets old. The model starts making mistakes and has to be updated with the new data.
- Regression
A task where the answer is a number instead of a class. How many steps, how many degrees, how full.
- MAE
Mean absolute error. The average of the differences between the estimate and the true value, ignoring the sign. It measures the error of a regression in the same unit as the answer.
- Forecasting
Estimating the next values of a series by looking at the past ones, for example how much battery will be left tonight. It's a regression where the input is the past.
6 questions that weren't in the lessons.
They bring the course's lessons together. After each answer there's the explanation and the place to read again. The review doesn't change the map.
-
Your slowest gesture, a big circle, lasts three seconds. The window is two seconds long. What happens?
For a gesture, the window has to hold the whole thing. Time your slowest gesture and pick a window a little longer, here more than three seconds. Explained in 04.01, chapter 4 →
-
You test the gesture model in two ways. With tests picked at random from everyone's examples, it gets 95 percent right. Leaving out one person at a time, 80. Which number do you tell the people who will use the app?
With random tests, the model has already seen gestures from the same people, and the score goes up. Leaving one person out measures what happens with someone new. Explained in 04.02, Go deeper →
-
The washing machine model raises the alarm every time someone bumps into the cabinet. What do you change?
A bump makes a single window strange, while a real fault stays. Asking for several strange windows in a row removes the isolated alarms, just like with live gestures. Explained in 04.03, chapter 5 →
-
Your rhythm model has an MAE of 12 beats. The fake model, which always answers with the average, has 13. What do you make of it?
The fake model doesn't even look at the input. The gap between the two tells you how much yours learned, and one beat is almost nothing. Explained in 04.04, Go deeper →
-
Every day you save the average strangeness score of the washing machine. You want to estimate what it will be a month from now. What task is that?
You estimate a future value by looking at the past values of the same series. It's still regression, and it's called time series forecasting. Explained in 04.04, chapter 6 →
-
You want to know whether you're walking slowly, at a normal pace or running. What should you train?
When the answer has only a few possible values, a classification is easier to train. Regression is for when the values in between really matter, like the exact beats of a rhythm. Explained in 04.04, chapter 1 →