Category Archives: Uncategorized

Winter Innovation – Hot Water Evaporator?

I bought an arctic air cold water evaporator. It work well for near field personal cooling.

So even though its summer time. I wonder. Could you heat a room or create near field warming. This with hot water instead of cold.

Perhaps the water temperature only need to be 35 degrees celsius and not boiling. Since this would break the device. Then the evaporation cools it down a bit. But it should still be warm. A 10 degree drop I guess.

Could be the most efficient way to save money on electricity. USB powered heater. But I think there needs to be a special version of the device.

Machine Learning Math – Solution To A Stiff Differential Equation

Solved a stiff diff equation using the chainer python module. Taken from wikipedia. The way I did it was to iterate with rng() random as the input with a while loop. That is. Iterate until the model(rng()) outputs the correct number.

Some problems were that dt could not be too small or it would get zero for update(). So I put it at 0.01 which I think is pretty high.

The loop for the initial model outputs.

y0 = model0(rng()) # previous value, y0
y1 = model1(rng()) # next value, y1
yp = (y1-y0)/dt # derivate
loss_eq = F.mean_squared_error(yp,-15*y0) # first error loss equation
loss_initial = F.mean_squared_error(init,y0) # inital error loss
loss = loss_eq + loss_initial
model0.cleargrads()
model1.cleargrads()
loss.backward()
optimizer0.update() # update each model0, model1
optimizer1.update()

source: diff eq y0 = model(rng),y1 = model(rng),dt = model(rng)(1)

dt could not be a model since it was so picky. Had to set it dt = 0.01

Machine Learning Idea – Failure Prediction Of Deep Learning Models(). The Practical Halting Problem?

“In computability theory, the halting problem is the problem of determining, from a description of an arbitrary computer program and an input, whether the program will finish running (i.e., halt) or continue to run forever.” – wikipedia

I think this is possible and quite important. If we rewrite the halting problem to a practical, still useful problem. Then “from a description” becomes from a small number of initial steps.

Then the problem could be fed to another machine learning model(). With indata like the model parameters, data from inital step calculations. Like the input, output and error.

Then the idea is that the practical halting model() will output a number indicating the probability of success.

If this works. The lessons is that even if something seems impossible you can rewrite the rules a bit but still get something useful back. Do we need to change the approach of the model().

Super Spline – Machine Learning Spline Replacement model() For Specific Types Of Curves

The idea is simple.

Why not invent something of a replacement for cubic splines. I’m thinking of using a machine learning model() as smart interpolation.

The idea is that a cubic spline might resemble one type of events where as a ML model() could learn to interpolate any type of curve.

So it would work a little like super resolution. You have a limited numbers of samples and you apply the ML-spline-model(). It then looks more like the true signal.

Summer Heat Idea – Air Cooler On A Powerbank Battery

Why battery? Running on battery means you can move the air cooler to where your at. Then using a foldable solar charger to charge the battery. Means that those without reliable power or no power at all also could have a small air cooler. For rural Africa perhaps.

The USB powered Artic Air had three speed settings. So the running time with a powerbank will vary. It came with a usb charger but it could run on computer usb. So the manual said. So I assumed a powerbank could do the same job. The powerbank was rated at 2.1A.

Anyway 30+ degrees at home without an air conditioner. Just fans. Is hard. This air cooler worked by near field personal cooling. Sitting infront of the cooler is what I mean.

Solar Energy For Sweden Idea – Large Foldable Solar Park On Wheels. Switch Place To Place. Follow The Sun On Electric Trucks

I think I imagined a new business. The energy truck driver.

So the idea is for a large foldable solar panel park on wheels. Where the operator goes to place to place on demand. Partly because the energy grid is not developed. So it could be for emergencies, refugee camps, black outs etc. Or maybe just as a business for earning money.

Anyhow this is a very important idea for a resilient world.

 

Physics Speculation – Inverse Reasoning. Singularities As A Energy Storages. High Amplitude

You split a problem to solve it more easily. Therefor I wonder if the  universe is really a big energy problem. Splitted into many small solvable objects. I mean. Would this mean it stores energy in many small singularities. Since a singularities could have high value. Like divided by zero. This would be a managable way to store lots of energy in a safe way.

So the problem of the universe is safe energy storage?

Machine Learning Innovation? – Sort Data Into Multiple Model Groups From The Model With The Best Confidence. Triangle Model System.

I wonder if this learning technique could work. The idea is to split the data into two or more model() groups. Each belonging to a separate model().

So the way I would test this would be to look at some sort of ?confidence score for each classification the two models could do. Like in the long vector binary encoding. Where you want 1.0 and the rest to be zeros.

So then the idea is to move the sample data into the model group with the highest confidence. This way the two models compete to have data in its group.

Then if I get 100% confidence for each model for training data within its group. Then the sample data transformation output from the test set would be selected from the model with the highest confidence score.

My first try (jupyter notebook) Sort Data into Model() Groups

98.11% Accuracy score

Impressive Feature – Using Machine Learning Super Resolution ( Down Scaled Images ) To Fix Blurry Images ( a little )

I managed to fix some photos that were blurry simply using a super resolution machine learning algorithm.

The idea is simple. Down scale the photo. Until you don’t see the blurriness. Then apply super resolution. Its as simple as that. The result is not any way perfect but its better.

Super resolution should be in gimp I think.

Video Or Image Compression Idea – Example. Optional JPEG Algorithm + Machine Learning Blockiness Layer Removal

Since machine learned super-resolution worked. Why should not do the same training for images with visible blockiness.

The idea is to degrade or hard compress images using some image compression algorithm to blockiness level. Use those images as the training and then as the target use the original images.

If this works than it would be possible to compress images to near ?the maximum.

The maximum ?would be unsupervised labeling of small objects in the image to text form and then machine learning to imagine what the image would look like.

So now you have a machine learning layer on top of chosen algorithm.

Perhaps this could work on youtube also. It would be possible to compress video on quality, not just on resolution. Then use the machine learning layer to correct the blockiness and resolution.

I think this is the way to do it. You have a standard algorithm in the back and a machine learning layer on top of it.

Inventions Idea – Online Volunteer Engineering – Then Order A Prototype Online

As the weather gets very hot. One invention that is missing is an air cooling device. Run by solar. Small enough that it does not cost that much.

For such climate inventions to be built I wonder if we need to connect engineers online.

So the idea is for an online engineer volunteering site with connected online prototype manufacturing site.

Just because we need it.

Machine Learning Idea – Server Heat Dependent On The Type Of Layer Used? LSTM

I wonder. My little laptop goes super hot every time I use something like LSTM. I guess it could be the cache. However. For many processors in a room this could make a substantial difference I guess.

So the idea is for optimize algorithms with machine learning for low energy consumption or heat dissipation. Could there be a “lagom” speed for the amount of heat?