Wikipedia has a full explanation
showing how feedback makes
System output in the diagram below lock in on a reference point.
With that as background, we will focus on how this
Biofeedback app works to put you on a
"Glide Path"
to land your BMI on your goal following a natural
Decay Curve
that avoids rebounds.
At the right hand side, System output is body
mass,1
as determined by the
System (our body)
piling upii
calories (
System input) that it does not burn.
The
Sensor is a
combination2
of two mathematical
An EMA, rejecting some of the noise
while keeping 100% of the slope.
Then Least Squares Fit of a line through the
remaining noise.
filters
built into this program and an ordinary
Notice the feedback - you want to control its reading.
To do so, you respond to a Red/Green warning signal.
scale.
The
Measured output is a
projection3
of the
Granted, you do see the scale weight before recording it.
The secret ingredient is to focus on the projection.
scale
reading and the
rate ofii
descent toward the
Reference point
(goal
Type it in at the bottom to tell the program when to switch
back and forth between red and green.
you set
).
This app is an
Also called Bang-Bang control.
This was chosen to allow mental recovery time.
On/Off
control system - like a
The program tells you when you need a new surge of
will-power.
thermostat
and furnace.
The
Measured error
is presented as "turn on the will power"
or "dropping
fast enough
".
The remaining
item is you4
- the
Controller -
who closes the loop.

Tackling one
Burn-out is less of a problem.
Habits allow periods of low effort.
habit
at a time will keep the BMI
Like the furnace, will-power shuts down periodically.
Like a house, the body and the habits smooth out the result.
heading
toward the goal if we keep the signal mostly green.
Control system jargon calls that
"
locking in".
This method has a subtlety - the
rate of descent
is what the filters extract (the "m" in Y=mX+b
of linear regression), and if one "diets" too
vigorously, the signal turns red again.
Thus the speed with which one heads toward goal is controlled
to follow a natural
decay curve;
getting to goal faster than a harsh diet would.
FOOTNOTES:
1
The output of a system is what you want to control,
which in this case is the scale reading.
The System Input, to its left, is calories in food.
For the body, there is a delay of about a year in
losing calories that have been consumed.
A comparable system would be a furnace, house
and thermostat,
in which case the system output would be temperature
and the system input would be energy in e.g. natural gas.
For a furnace, the walls and interior
of the house slow the loss of heat.
The temperature stays comfortable
after the furnace shuts off and before the
thermostat turns it back on.
2
The sensor is an ordinary scale, but functionally, it is
only part of a sensor subsystem which includes two filters
in the app and the involvement of the user to type in data.
Note that the input of the sensor is the output of the
system.
The first mathematical filter is an EMA, which rejects 97%
of the scale fluctuations while keeping 100% of the slope.
Then a Least Squares Fit of a line through the remaining
noise reduces it and extracts the slope, which is the rate
of drop in scale readings.
These filters are in the program, and rely on the user to
read a scale and type in the data.
One more step is required to make the user an integral part
of the feedback loop; the extracted slope is used to
calculate the projected weight and display it.
3
The user is shown a two-state comparison of the future
weight and the goal - either green for "no new habit
needed" or orange-red for "time to re-focus".
Diets cause burn-out by demanding extra effort until the
goal is met. Then the user must put more effort into
developing a maintenance routine to keep the weight off.
People naturally relax until things get really bad again
and "yo - yo dieting" rsults.
This sensor design guides one's habits from the
initial fast drop smoothly to staying on the goal.
4
Feedback control systems deal with information, collected
by the sensor and presented to a comparator (shown by the
plus and minus by the arrows to the left above). In this
case the information processing is not done by a thermostat,
but rather by the user of the app.
Normally one focuses on the day-to-day changes in
scale weight, but the use of heavy filtering and the
shift from actual weight to rate of descent leads one
to avoid discouragement by showing success all the way down.
ii
ASIDE: For those who set up PID controllers, the body's
accumulation of unburned calories is an integral process.
The derivative from Linear regression changes the loop
gain back to Proportional. In the case of heating a house
there is no derivative feedback so there is no overshoot
like this app exhibits.