Content of the material
Resolution Vs. Precision Vs. Accuracy
When considering scales, three main qualities determine whether they are working properly, the first of which is resolution. Resolution is the value that tells the user how close the scale can read to the object’s true weight. For example, a scale may have a resolution of one gram. What this means is that if an object weighs 20 grams, then the scale will give a reading for the object up to one gram either way from the object’s true weight.
Precision differs from accuracy because accuracy is needed once, precision is needed all of the time. If you weigh the same object repeatedly but get different measurements each time, then the scale has low precision. On the other hand, if you get the same reading every time you weigh the object, then you have perfect precision.
However, precision does not equate to accuracy. For example, if the resolution of the scale is low, even with perfect precision, it will not give you the true weight of an object. In order to have high accuracy, the scale must have both high resolution and high precision.
In fact, it’s possible to create an extremely precise scale that will never return the true weight of an object laid on it. In the past, some retail scale vendors would actually go around bragging about the quality of admittedly bad scales. The reason was, simply, precision on them was really high in spite of all the other metrics being low.
Technically, those kinds of claims don’t constitute false advertising since the person making them isn’t claiming anything that’s untrue. Rather, they’re merely withholding information to make their products look better than they are.
Don’t fall into that kind of trap. Make sure that you’re working with a legitimate vendor that has experience in the field. While you’re at it, learn a little more about why scales may not be as accurate as they initially look to be.
Testing Precision Equipment
Scales used in scientific, engineering, manufacturing and similar types of operations often require extremely high accuracy. For such testing, it is necessary to use certified calibration weights instead of more common objects like dumbbells.
A good method is to use three weights with a ratio of about 1:2:4. Also use a sum that represents a substantial percentage of the scale’s capacity. Start off by zeroing the equipment as in the process mentioned above. Then begin measuring weights to see whether they produce the same results consistently, as this will test for precision.
To test accuracy, weigh the calibrated weights in different combinations. Then see if the results represent the correct sums of the weighted objects. These procedures measure the linearity of the scales. A taring test determines whether the linearity tests continue to produce the same results when the scale resets to zero or to a non-zero calibration.
When conducting testing of high accuracy scales, use a room with a controlled temperature. For some equipment, the internal temperature of the equipment is also important. Additionally, weighing equipment can be sensitive to static electricity and radiofrequency or electromagnetic interference.
The term ‘drift’ describes inaccuracies caused by environmental factors like temperature and static electricity. For equipment used in conditions in which these variables are not controllable, then it will be necessary to measure the sensitivity drift of the scale.
Manufacturers provide temperature ranges for their equipment along with calibration certification in some cases. Note that calibration certification not might be valid if the equipment comes from a distant location. For devices that require high accuracy, even differences in geomagnetism and barometric pressure can result in measurement differences when compared to the equipment calibration location.
The frequency of calibration will depend on how important accuracy is to the operation. Another important factor is the conditions that might cause the equipment to lose accuracy. Some organizations may recalibrate weighing systems on a monthly basis. Others conduct testing at the start of each shift or even before each use of the equipment.
1. Confusion Matrix
A confusion matrix is a table that helps visualise the performance of a classification model.It can be used to calculate Precision,Sensitivity(aka recall),Specificity and accuracy.
Definition of the Terms:
- True Positive (TP) : Observation is positive, and is predicted to be positive.
- False Negative (FN) : Observation is positive, but is predicted negative.
- True Negative (TN) : Observation is negative, and is predicted to be negative.
- False Positive (FP) : Observation is negative, but is predicted positive.
Precision = TP/(TP+FP)
Testing a Scale
So how can you test a scale?
Before you begin, you want to make sure you’ve set the stage properly. That is, make sure your bathroom floor (or whatever other surface you’re using) is hard and flat. Using a scale on carpet or an uneven surface will throw off the sensors, both mechanical and digital. Pick up the scale and clean beneath it; even small bits of debris can have an impact.
You also may need to consider environmental factors. Scales may be subject to something called “drift”. Drift is when your scale works fine, but the measurements aren’t quite the same day to day. Why not? Things like temperature can affect the resilience of the materials used to construct the scale. If it’s colder, springs and load gauges might not flex the same amount, and can give different readings.
As far as testing your scale goes, here are some steps you can take.
First, make sure you zero out the scale properly. A digital scale will have some means of zeroing it out, whether it’s an action like stepping on and off, or a button you can press to reset it. Refer to the instructions on the scale, or if you don’t have the instructions, look for them online. You can usually find some instructions if you can Google search the model number of the scale, and possibly even a PDF of the original instruction booklet.
For mechanical scales, there may be something simple like a dial you can turn to make sure it reads zero before you step on it. For other scales, you may need to open it up and turn something internally to make the same adjustment. Be careful if you do this; the last thing you want to do is break your scale.
Incidentally, you want to calibrate or zero out your scale every time you use it. With proper calibration, home scales can be perfectly accurate for weight monitoring purposes. If you move the scale, though, it needs to be recalibrated. In general, you want to calibrate it once every month or two as well, just in case. When in doubt, zero out!
For electronic scales, it might also be worthwhile to replace the batteries. When batteries get low, the device will still work, but might not be as accurate.
Second, test for consistency. Step on your scale, get a reading of your weight, and step off. Repeat this process four or five times, and note if there’s any variance in your weight. Make sure you’re stepping on it the same way, and that you’re not bouncing, shifting, or otherwise adjusting your weight distribution as you test it.
If your scale is consistent, great! You can move on to the next step. If it’s not consistent, you may have an issue with the internal elements of the scale. This is usually a symptom of something breaking, whether it’s the load gauge in the digital scale breaking down, or the pinion sticking inside the mechanical scale.
In these cases, you can potentially open up and repair the scale, but it’s generally easier to just go buy a new one. Old scales break down, and it’s usually more trouble than it’s worth to repair them.
If you’ve determined that your scale is consistent, then you can try to test it with something with a known weight. We like to use small hand weights, something like these, which have a known weight written on the side of them. You want to use something around 10lbs or so. Some scales don’t “activate” for lower amounts of weight, because of how they’re calibrated internally.
If you put a known 10lb weight on your scale and it shows something other than 10 pounds, you know it’s not necessarily accurate. You can try to recalibrate it again, but chances are it’s just a margin of error for how the scale works.
You can also take more than one known weight – like two of those dumbbells – and weight them at the same time. If 1x10lb weight measures in at 9 lbs, and 2x10lb weights measure in at 19lbs, you know it’s within 1 lb of accurate.
That’s one of the more difficult things to accept. The fact is, most bathroom scales aren’t going to be too terribly accurate, at least not to a sub-pound level. They generally only have a couple of sensors, and their tolerances aren’t great, so they will be accurate within about 2-4 pounds. Unfortunately, when you’re trying to chart your weight loss over time, being “off” by a few pounds isn’t great.
How Often Should You Calibrate Your Scales?
How often you calibrate your scales depends on a few different factors — manufacturer’s recommendations, how often you use the scales, the environment they’re in, and how essential an exact weight is to your business. Some are calibrated once per month, others are only calibrated once per year, while some are even spot checked daily for accuracy.
Usually, after considering these factors, it’s determined that somewhere in the middle —certified calibration once per quarter with a weekly user spot check — is ideal to ensure quality control of scales that are used fairly often. However, your need may vary based on the information below.
Always consult the manufacturer’s recommendations first. Manufacturers should have a recommendation for the frequency of calibration, and since all scales are made differently, it’s usually safe to assume the manufacturer knows best.
If your scales are used multiple times throughout the day, every day of the week, normal wear and tear will occur faster than in scales that are used a few times per week. Therefore, if you use these scales more frequently, they should be calibrated more frequently — perhaps monthly.
The surrounding environment also plays a role. For example, if your scale is in an area that contains dust, fluids, or other substances, your scale could have a buildup of these substances that interfere with the performance of the scale. Or, if you have the scale located in a place where there are vibrations, static electricity or mechanical shock, scale accuracy can suffer.
In general, you should also consider a more frequent calibration to ensure these substances and other work environment factors are not getting in the way of getting a precise weight.
Finally, consider how important an accurate weight is to your business. If your company can’t afford to have even the slightest inaccuracy in weight, it’s likely that more frequent calibrations will be necessary. It may be easy to understand the importance of accuracy when blending pharmaceutical ingredients where a mistake in weighing a single batch could easily exceed $100,000 in cost. But what about rocks and stones from a quarry where a truckload might only be worth a few hundred dollars. Would a small error of 1% make much of a difference? Maybe not for that one load, but when you take into account that the same scale also weighed 100 other trucks each day, you could be looking at a loss of over $75,000 if you wait a whole year to calibrate that scale.
Calibrations are done at a variety of frequencies, depending on a variety of factors. It’s important to note that if your equipment is frequently calibrated and almost always needs adjustment or repairs, it could be a sign of a bigger issue.
While the frequency of calibrations depends upon use, the rapid deterioration of accuracy warrants additional troubleshooting.
Once you’ve determined the appropriate frequency for your scales, you should look into signing a service plan with a calibration company. Many companies, like Precision Solutions, will work with you to learn more about your process and then create a maintenance plan tailored to you and your equipment
The Scale of Your Data Matters
Deep learning neural network models learn a mapping from input variables to an output variable.
As such, the scale and distribution of the data drawn from the domain may be different for each variable.
Input variables may have different units (e.g. feet, kilometers, and hours) that, in turn, may mean the variables have different scales.
Differences in the scales across input variables may increase the difficulty of the problem being modeled. An example of this is that large input values (e.g. a spread of hundreds or thousands of units) can result in a model that learns large weight values. A model with large weight values is often unstable, meaning that it may suffer from poor performance during learning and sensitivity to input values resulting in higher generalization error.
One of the most common forms of pre-processing consists of a simple linear rescaling of the input variables.
— Page 298, Neural Networks for Pattern Recognition, 1995.
A target variable with a large spread of values, in turn, may result in large error gradient values causing weight values to change dramatically, making the learning process unstable.
Scaling input and output variables is a critical step in using neural network models.
In practice it is nearly always advantageous to apply pre-processing transformations to the input data before it is presented to a network. Similarly, the outputs of the network are often post-processed to give the required output values.
— Page 296, Neural Networks for Pattern Recognition, 1995.
Scaling Input Variables
The input variables are those that the network takes on the input or visible layer in order to make a prediction.
A good rule of thumb is that input variables should be small values, probably in the range of 0-1 or standardized with a zero mean and a standard deviation of one.
Whether input variables require scaling depends on the specifics of your problem and of each variable.
You may have a sequence of quantities as inputs, such as prices or temperatures.
If the distribution of the quantity is normal, then it should be standardized, otherwise the data should be normalized. This applies if the range of quantity values is large (10s, 100s, etc.) or small (0.01, 0.0001).
If the quantity values are small (near 0-1) and the distribution is limited (e.g. standard deviation near 1) then perhaps you can get away with no scaling of the data.
Problems can be complex and it may not be clear how to best scale input data.
If in doubt, normalize the input sequence. If you have the resources, explore modeling with the raw data, standardized data, and normalized data and see if there is a beneficial difference in the performance of the resulting model.
If the input variables are combined linearly, as in an MLP [Multilayer Perceptron], then it is rarely strictly necessary to standardize the inputs, at least in theory. […] However, there are a variety of practical reasons why standardizing the inputs can make training faster and reduce the chances of getting stuck in local optima.
Scaling Output Variables
The output variable is the variable predicted by the network.
You must ensure that the scale of your output variable matches the scale of the activation function (transfer function) on the output layer of your network.
If your output activation function has a range of [0,1], then obviously you must ensure that the target values lie within that range. But it is generally better to choose an output activation function suited to the distribution of the targets than to force your data to conform to the output activation function.
If your problem is a regression problem, then the output will be a real value.
This is best modeled with a linear activation function. If the distribution of the value is normal, then you can standardize the output variable. Otherwise, the output variable can be normalized.
Want Better Results with Deep Learning?
Take my free 7-day email crash course now (with sample code).
Click to sign-up and also get a free PDF Ebook version of the course.
Can bathroom scales lose accuracy?
Yes. Age, wear and tear can take their toll on your bathroom scale.
Analog scales may be more prone to damage—for example, if the levers bend out of shape or if dirt gets into the springs.
Digital scales may be less likely to become contaminated. If you have a smart scale that’s connected to an app, make sure you install any suggested updates to keep your scale running smoothly.