Modeling with Exponential Functions
Introduced| English | Meaning |
|---|---|
| exponential model/ˌekspəˈnenʃl ˈmɒdl/ | A model in which equal input increments multiply the output by a constant factor. |
| initial value/ɪˈnɪʃl ˈvæljuː/ | The starting value of a quantity, function or system at a specified initial point or time. |
| growth factor/ɡrəʊθ ˈfæktə/ | growth factor |
| decay factor/dɪˈkeɪ ˈfæktə/ | decay factor |
| baseline/ˈbeɪslaɪn/ | baseline |
| natural base/ˈnætʃərəl beɪs/ | The number e ≈ 2.718, the base of natural exponential and logarithm functions. |
| exponential regression/ˌekspəˈnenʃl rɪˈɡreʃn/ | Fitting an exponential model to observed data. |
| residuals/rɪˈsɪdʒuːəlz/ | An observed value minus the value predicted by a model. |
The first measurement is not always the start
- A culture has $180$ cells at $t=2$ hours and $1440$ cells at $t=5$ hours. If an exponential model 指数模型 is appropriate, what happens in one hour, and what does it predict at $t=0$?
- The count multiplies by $8$ between the measurements, but that change takes three hours. A model must connect the multiplier to the length of its time interval.
Say what the parameters mean
- In $N(t)=ab^t$, the initial value 初始值 is $N(0)=a$. It is the value at the chosen time origin, not automatically the first recorded observation. Here $N(2)=180$, so $a$ is still unknown.
- For positive counts, use $a>0$ and $b>0$. Two positive observations at distinct times determine one model of this form; they do not prove that exponential growth is the right description of the situation.
In the unshifted model $y=ab^t$, the coefficient $a$ is the output when…
Difficulty 1/5 · 100 coins + a small bonus on first completion today
At $t=0$, $b^0=1$, so $y=a$ in this unshifted model. The first observed time need not be zero.
Divide the two equations
- Substitute the two observations into $N(t)=ab^t$, then divide the later equation by the earlier one. The coefficient $a$ cancels, leaving the multiplier for the whole time gap.
- Take the positive cube root to obtain the multiplier per hour. Then substitute back into either equation to find $a$.
The model is $N(t)=45(2)^t$. It reproduces both measurements: $45(2)^2=180$ and $45(2)^5=1440$.
A culture grows from 180 to 1440 cells over three hours. What is the multiplier across the whole three-hour gap?
Difficulty 2/5 · 140 coins + a small bonus on first completion today
The full-gap multiplier is $1440/180=8$. This is not the one-hour multiplier.
Attach each factor to its interval
- The growth factor 增长因子 is $2$ per hour, so the count increases by $100\%$ each hour. Across three hours the factor is $2^3=8$, an increase of $700\%$, not $800\%$.
- If $s$ counts three-hour periods from the same origin, then $t=3s$ and $N=45(8)^s$. Changing the input unit changes the base; it does not change the predicted count at the same instant.
A positive exponential population multiplies by 8 over three hours. What is its multiplier per hour?
Difficulty 2/5 · 140 coins + a small bonus on first completion today
$b^3=8$, so the positive hourly factor is $b=2$.
Predict and check the time origin
- At $t=4$ hours, $N(4)=45(2)^4=720$ cells. Equivalently, start at the first observation: $N(t)=180(2)^{t-2}$. Both formulas describe the same model.
- The model's $45$ cells at $t=0$ are inferred, not measured. For a different culture modeled by $P(t)=100(3)^t$, the count at $t=2$ is $900$; each coefficient refers to its own chosen start.
A colony follows $P = 100 \cdot 2^{\,t}$ (with $t$ in hours). What is $P$ at $t = 3$?
Difficulty 2/5 · 140 coins + a small bonus on first completion today
$P = 100 \cdot 2^3 = 100 \cdot 8 = 800$. Plug the input into the model to predict.
An exponential model doubles each hour and gives 180 cells at hour 2. What does it predict at hour 0?
Difficulty 2/5 · 140 coins + a small bonus on first completion today
Reverse two hourly doublings: $180/2^2=45$. This is an inferred value at the chosen origin.
Convert percentages into factors
- For a positive quantity, a fractional increase $r$ per period gives $b=1+r$. Thus $20\%$ growth gives $b=1.20$; a $15\%$ loss gives the decay factor 衰减因子 $b=1-0.15=0.85$.
- A positive base below $1$ represents decay toward zero in $ab^t$; a base above $1$ represents growth when $a>0$. A negative base does not define a real-valued exponential for every real time.
If a quantity grows by $20\%$ each period, the growth ____ (the base $b$) is $1.2$.
Difficulty 1/5 · 100 coins + a small bonus on first completion today
A $20\%$ increase multiplies by $1 + 0.20 = 1.2$ — that multiplier is the growth factor.
A quantity that loses $15\%$ each period has a decay factor (base) of…
Difficulty 1/5 · 100 coins + a small bonus on first completion today
Losing $15\%$ leaves $85\%$, so you multiply by $1 - 0.15 = 0.85$ each period — the decay factor.
Remove the background before taking ratios
- A room stays at 20°C while a drink cools. At hours $0,1,2,3$, the drink's temperatures are 100°C, 60°C, 40°C and 30°C. The raw temperature ratios are not constant.
- Subtract the baseline 基准值 of $20$: the temperature differences are $80,40,20,10$. These halve each hour, so the model is $T(t)=20+80(0.5)^t$, approaching room temperature rather than zero.
Using the readings at $t=0$ and $t=2$, the ratio of differences is
For the shifted model $c+ab^t$, the initial output is $c+a$, not $a$. The baseline must be known or separately estimated; two readings alone do not determine all three parameters $c,a,b$.
Build the curve from its observations
The culture has 180 cells at hour 2 and 1440 at hour 5. Compare its three-hour factor with its hourly factor, then examine cooling above a baseline.
A cooling model is $T(t)=18+62(0.5)^t$. What temperature in degrees Celsius does it give at $t=0$?
Difficulty 2/5 · 140 coins + a small bonus on first completion today
$T(0)=18+62=80$. In $c+ab^t$, the initial output is $c+a$, not just $a$.
Shift the input as well as the output
- Count elapsed time from the $t=2$ cooling reading by setting $u=t-2$. At $u=0$ the drink is $40^\circ\mathrm{C}$, which is $20^\circ\mathrm{C}$ above the room.
- The same physical model becomes $T=20+20(0.5)^u$. Check: $u=1$ means original time $t=3$, giving $30^\circ\mathrm{C}$. Track the time origin, input unit and baseline when changing a formula.
Write the same model with base e
- The natural base 自然底数 is $e\approx2.718$. Since $b=e^{\ln b}$, the forms $ab^t$ and $ae^{kt}$ agree when $k=\ln b$; an added baseline stays outside the exponential.
- The culture model is $N=45e^{(\ln2)t}$. Here $k\approx0.693$ per hour, while the ordinary one-hour percentage increase is $100\%$. The exponent coefficient $k$ and the fractional increase $b-1$ are different quantities.
The model $N(t)=10e^{(\ln1.2)t}$ has what ordinary percentage increase over one unit of time?
Difficulty 1/5 · 100 coins + a small bonus on first completion today
$e^{\ln1.2}=1.2$. The one-unit fractional increase is $1.2-1=0.2$, or 20%; it is not the exponent coefficient $\ln1.2$.
Use more observations to test the model
- With noisy measurements, use an exponential regression 指数回归 to fit the data with technology. Some observations may lie on the fitted curve; others may not. Check the resulting predictions against all the measurements.
- A log-linear fit regresses $\ln y$ against time when $y>0$, then converts back. For a known baseline $c$, apply this to $\ln(y-c)$ only when $y-c>0$. Compare observed minus predicted values, called residuals 残差, rather than trusting the equation alone. The next lesson compares their sizes and patterns.
Passing exactly through two positive observations proves an exponential is the best model for all later times.
Difficulty 1/5 · 100 coins + a small bonus on first completion today
Two points determine a candidate unshifted exponential, but cannot establish its contextual suitability or long-range reliability. Use additional observations and context.
State where the prediction is useful
- Exponential population growth can be a short-range model while nutrients and space are available. A fixed-room-temperature cooling model assumes that the room stays at the stated baseline and that the cooling conditions remain similar.
- Earlier or later inputs are not automatically meaningless, but predictions outside the observed interval need extra justification. In particular, fitting two points exactly does not establish that a distant forecast is reliable.
The first culture measurement is at hour $2$. A model can estimate hour $0$, but that estimate depends on the same growth pattern applying before the measurements. Label an inferred starting value clearly.
Find the factor over the full input gap, convert it to a factor per input unit, and solve for the starting coefficient. For $c+ab^t$, first work with the difference $y-c$. Check both observations, additional data and the contextual limits of the model.
For the positive-quantity model $y=ab^t$, with $a>0$ and $b>0$, select all true statements.
Difficulty 2/5 · 140 coins + a small bonus on first completion today
The first three statements hold for this unshifted positive model. Forecasts beyond the observations require contextual justification; a formula alone does not establish reliability.
Write a prediction with its assumption
- A colony follows $P(t)=50(3)^t$, with $t$ measured in hours from the model's start. Calculate $P(3)$ and explain how the coefficient and the factor are used.
- State one assumption needed to use this formula for a later forecast. Distinguish a model prediction from an observation.
A colony follows $P(t)=50(3)^t$, with $t$ in hours from the model start. Find $P(3)$, explain the roles of 50 and 3, and state one assumption needed for a later forecast.
Calculate from the stated time origin, identify the interval attached to the base, and name a condition needed to extend the model.
Using $P(t)=50(3)^t$, what population does the model predict at $t=3$ hours?
Difficulty 1/5 · 100 coins + a small bonus on first completion today
Use this question's model: $50(3)^3=50(27)=1350$. A prediction from a different formula does not answer the question.