In 2015 New Year’s resolution , I said that I want to lose 10–15 pounds by June 2015. In Weight-loss progress report, Second weight-loss progress report, and Third weight-loss progress report I provided monthly updates.
April was keeping to the same diet as the previous three months: only raw fruits and vegetables for lunch on weekdays, and slightly reducing how much I ate in the evenings. We probably had more meat this month than usual, and I did have some snack foods and desserts that I haven’t been eating in previous months, but the weight loss continued at about the same pace as before, perhaps even slightly faster:
The exponential fits to my weight loss that I tried in March turned out to have no predictive value—the straight line predictions of my weight were consistently more accurate, despite having fewer parameters to fit to the preceding data. Daily updates to the exponential fit made it closer and closer to the straight line, so I gave up on doing them—my weight loss was not behaving like an exponential decay.
I have now lost the 20 lbs that was my most ambitious goal, and I’ll need to switch my control algorithm to try to maintain a constant weight, rather than lose a pound a week. I don’t want to have huge rebound in weight from ending the reducing diet, nor do I want to keep getting skinnier—I’m back in the weight range I was about 20 years ago when I was 40, and I’d like to stay there.
In coming up with a control algorithm, I need to take into consideration how much fluctuation there is in my weight (I don’t want to be changing plans all the time because of noise) as well as how fast my weight can change as a result of deliberate actions. It looks to me like my weight fluctuates about ±2 lbs around the trend line, so a target range that is 5lbs wide is probably going to be as tight as I can control my weight.
I’m thinking of using a two-state control algorithm with hysteresis and time delay:
- Normal diet: If my weight is ≤ 156 lbs for 2 days in a row, I’ll eat lunches with a fair number of calories (from the taco truck or sandwiches).
- Reducing diet: If my weight is ≥ 158 lbs for 2 days in a row, I’ll use my current reducing diet (only raw fruit or vegetable for lunch).
I may add a couple more transition points at the edges of the range (extra desserts if I fall below 155 pounds, more attention to eating less in the evening if I go over 160lbs), but I’m hopeful that the control algorithm won’t hit those boundaries very often. If I see a periodic fluctuation in my weight, I may switch to faster control (not weight for two days above the high threshold or below the low threshold).
Right now I’m still on the reducing diet, but I may hit the first transition point in the next week and start eating lunches with more calories again.