Have you considered a positive feedback capacitor with the op-amp, to try to make an oscillator?
Then the illumination on your LED will affect the phase of the oscillator. By trying to make it oscillate at ~10Mhz, and then sampling the data, you should avoid the clipping problem, and you'd have a lot more data to use to potentially extract interesting things with clever math.
Even if the oscillator runs at 10Mhz, you don't have to sample at 10 Mhz - you could sample at a lower frequency (which must be clock cycle accurate) and make use of aliasing to still measure phase shifts very accurately.
You're totally going to measure the power supply, temperature, radio signals and a whole host of other things at the same time, but with the right signal processing/modulation you should be able to extract the effects you need.
Here's a much better explanation of the Fourier transform:
When you took calculus, you learned that functions can be written in a kind of "Taylor series":
f(x) = a0 + a1 x + a2 x^2 + ...
However, the functions {1, x, x^2, ...} aren't unique. You can use any basis functions. Let's replace them with {Φ0, Φ1, Φ2, ...}. To find your "Taylor series" you need some way to measure how far off your approximation is. We call this an "inner product", and one common one is the integral, i.e.
<f, g> = integral of f̅g dx
If we can change our coefficients {ai} and get a better approximation, it means there is some basis function where
So, to get the best approximation, we just set the left and right sides of the equation equal. This is really easy to calculate if most of the <Φi, Φj> = 0 (and cancel out). We call a basis orthogonal if <Φi, Φj> = 0 except when i=j. For an orthogonal basis, we're left with
<Φi, f> = ai<Φi, Φi> --> ai = <Φi, Φi> / <Φi, f>.
One common orthogonal basis are the Legendre polynomials, which are the same as {1, x, x^2, ...} with the Gram-Schmidt process applied to them. We're not really going to discuss those here. Another common one are sines and cosines. In trigonometry you learned that
e^{ix} = cos(x) + i sin(x),
so we can instead use the basis
{..., e^{-2ix}, e^{-ix}, 1, e^{ix}, e^{2ix}, ...}
Plugging this in gives the Fourier transform:
a_k = 1/2pi * integral from -pi to pi of e^{-kix}f(x) dx.
Foe reference the individuals and organizations pushing for the mass surveillance of the EU citizens under the guise of protecting the children are:
Ylva Johansson, the EU Home Office Commissioner. She's openly anti-encryption and has said she doesn't care about privacy or security concerns. She won't even meet with any group that disagrees with her.
Thierry Breton, the European Commissioner for Internal Market. He is working with Ylva Johansson and Thorn to pass Chat Control.
Monique Pariat, European Commission’s Director-General for Migration and Home Affairs
Catherine de Bolle, Europol Executive Director
Julie Cordua, CEO of Thorn.
Cathal Delaney, Former Europol employee who now works for Thorn.
Ruiz Perez, Senior former Europol official Fernando, who now is on Thorn's board.
Alan M. Parker, British billionaire, and founder of the Oak Foundation that bankrolls the fake charities lobbying for Chat Control.
Chris Cohn, British billionaire hedge fund manager and Google activist investor. He provides funding for anti-encryption lobbying in the North American and the EU.
Ashton Kutcher, Demi Moore. They try to whitewash Thorn's actions while lobbying on their behalf. The EU government let them bypass civil rights groups with their lobbying due to their fame. Other actors involved with Thorn can be found [here](https://en.wikipedia.org/wiki/Thorn_(organization)).
Ernie Allen, chair of the WeProtect Global Alliance, WPGA, and former head of the National Centre for Missing & Exploited Children, NCMEC, in the US. Part of the network of fake charities and corrupt organizations lobbying to ban encryption and privacy.
Sarah Gardner, former Thorn employee and now the head of the Heat Initiative. Part of the network of fake charities and corrupt organizations lobbying to ban encryption and privacy. She's focus on US lobbying.
Lily Rhodes, former Thorn employee and now the director of strategic operations at the Heat Initiative. Part of the network of fake charities and corrupt organizations lobbying to ban encryption and privacy. She's focus on US lobbying.
Maciej Szpunar, Polish Advocate General at the European Court of Justice. Wants to use the proposal for prosecuting copyright infringement.
Other individuals involved are: Margrethe Vestager, Margaritis Schinas, Antonio Labrador Jimenez, Douglas Griffiths, Javier Zarzalejos.
A non exhaustive list of the fake charities and corrupt organizations involved:
ECPAT, Eurochild, Missing Children Europe, Internet Watch Foundation, Terre des Hommes, Brave Movement, Thorn, Oak Foundation, WeProtect Global Alliance, Justice Initiative, Purpose
Organizations operating more in North America:
Hopewell Fund, Heat Initiative, Children’s Investment Fund Foundation
Finally, let's not forget that Ashton Kutcher, the darling of VCs, had to step down from his position at Thorn after submitting letters in support of their fellow actor and friend, a convicted rapist.
If he is willing to push for such privacy invasive measures in EU, he won't stop there, he will come for you in the US as well.
I think autonomous performance cars are using different but related techniques. I am not close enough to what the autonomous performance cars are doing to say where things are in terms of the state of the art but I think there is opportunity.
For Micromouse, there are several methods people use. One is to create a trapezoidal angular velocity profile while holding the forward speed constant. The trapezoidal profile parameters are determined through iterative simulation.
Another approach - the one I use - uses cubic spirals which is described in: Smooth Local Path Planning for Autonomous Vehicles by Yutaka Kanayama and Bruce I. Hartman. What is amazing about this technique is that it is closed form, is like four or five multiply and adds and executes in trivial time on (even) an 8-bit processor. For my latest entry, I have a more sophisticated scheme where I try to maximize the load on the tires and the lateral and longitudinal loads are asymmetric.
I think you can get very far with simulations and then trying it on a RC car and then on a real car.
A few years back, there was some amazing work that was done at Stanford where they developed tire models and a controller that could handle sliding modes.
I encourage you to explore because if nothing else, you will learn.
The gist is that all data types can be represented as programs, which have grammars, and ultimately can be converted to ASTs. The interpretation of the program is what leads to a value. I merge the ideas of state-based and operation-based CRDTs into one by enforcing an associative property, which is only possible if operations and states can both be built incrementally.
As an example, suppose we are applying a set of operations to a state:
state_0;
state_1 = op(state_0, arg_0);
state_2 = op(state_1, arg_1)
which can be represented as: op(op(state_0, arg_0), arg_1)
We can think of this as collapsing on the state side, in the sense that changes are encoded in the state. We can also collapse on the operation side, in the sense that changes are encoded in the operation:
op_0 = op(arg_0);
op_1 = op(op_0, arg_1)
which can be represented as: op(state_0, op(op(arg_0), arg_1))
I believe a CRDT that can collapse on either end will allow you to build conflict free replicated programs. Fundamentally what you are looking for is op(op(state, arg_0), arg_1) = op(state, op(arg_0, arg_1)), which is associativity.
What seems clever to me is the super cool hack of getting sensor data while the video data is captured, then training a model on realtime video and backward time-shifted sensor data; result is a network that can predict sensor data 30 seconds ahead of time with visual data. This is then used to feed the walking model. That's the sort of super 'dumb' thing ML folks can do now; that task would have taken many many person years of work before, now, I speculate thinking of the idea to testing a network would be under a week. Verry cool.
If you can afford it, following the intrinsic motivation may be exactly what you need to do.
Sometimes I'm in the position where I want or need to do something even though I'm dreading it or am not excited about.
A few things have helped me when that's the case:
1. Coaxing myself rather than trying to crack the whip.
- Finding what's interesting about the task
- Thinking through how it might be made interesting or easy
- Thinking through whether I can automate it away
- Starting on a tiny piece without committing to finishing it right away
- Lowering my requirements on quality for a first try to basically zero
- Developing a habit around the task with a reward like a streak or a check mark
2. Writing out my thoughts
- Why do I want to take on this task?
- Why don't I want to take on this task?
If I can drill down on both of those maybe I can come to some kind of resolution
Anyway, a lot of what works depends on what's broken in the current situation. I think that sometimes your desire to avoid something is right and sometimes it needs to be worked around. But in either case it can be at least heard with thoughtfulness and compassion. There's probably insight in the feeling that you don't want to miss out on.
On the other hand, I think it's a mistake to treat these feelings as the only thing either. Often even strong feelings are temporary and situational and getting better information or just letting time pass or building up a feeling of success from small wins will let the storm move on.
Some papers that are runnable on a laptop CPU (so long as you stick to small image sizes/tasks):
1) Generative Adversarial Networks (https://arxiv.org/abs/1406.2661). Good practice to have a custom training loops, different optimisers and networks etc.
2) Neural Style Transfer (https://arxiv.org/abs/1508.06576). Nice to be able to manipulate pretrained networks and intercept intermediate layers.
4) Physics Informed Neural Networks (https://arxiv.org/abs/1711.10561). If you're interested scientific applications, this might be fun. It's good exercise in calculating higher order derivatives of neural networks and using these in loss functions.
5) Vanilla Policy Gradient (https://arxiv.org/abs/1604.06778) is the easiest reinforcement learning algorithm to implement and can be used as a black-box optimiser in a lot of settings.
6) Deep Q Learning (https://arxiv.org/abs/1312.5602) is also not too hard to implement and was the first time I had heard about DeepMind, as well as being a foundational deep reinforcement learning paper .
I find this article odd with its fixation on computing speed and 8bit.
For most current models, you need 40+ GB of RAM to train them. Gradient accumulation doesn't work with batch norms so you really need that memory.
That means either dual 3090/4090 or one of the extra expensive A100/H100 options. Their table suggests the 3080 would be a good deal, but it's not. It doesn't have enough RAM for most problems.
If you can do 8bit inference, don't use a GPU. CPU will be much cheaper and potentially also lower latency.
Also: Almost everyone using GPUs for work will join NVIDIA's Inception program and get rebates... So why look at retail prices?
Then the illumination on your LED will affect the phase of the oscillator. By trying to make it oscillate at ~10Mhz, and then sampling the data, you should avoid the clipping problem, and you'd have a lot more data to use to potentially extract interesting things with clever math.
Even if the oscillator runs at 10Mhz, you don't have to sample at 10 Mhz - you could sample at a lower frequency (which must be clock cycle accurate) and make use of aliasing to still measure phase shifts very accurately.
You're totally going to measure the power supply, temperature, radio signals and a whole host of other things at the same time, but with the right signal processing/modulation you should be able to extract the effects you need.