Kamal

Keep it simple…

[Kamal @Kamall282]

I’m currently working on designing and simulating lightweight aircraft systems.

I will make a thread to explain each blocks and their functions.

Like & share and tag government bodies

[Kamal @Kamall282]

@1meajay @TOPNOTBOYGEM The sum of the interior opposite angles is equal to the extended angle outside (exterior)

[Kamal @Kamall282]

React if you can see this tweet

Elon father has shadowban my account 🙂‍↔️

[Kamal @Kamall282]

@arunaafeez @misty_abisola @chinedusolidbay

[Kamal @Kamall282]

figure

subplot(1,2,1)

scatter(YTest,YPred,50,'filled'); hold on

ref =linspace(min(YTest),max(YTest),100);

plot(ref,ref,'r--','LineWidth',2)

title('Neural Net')
xlabel('Actual'); ylabel('Predicted'); grid on

subplot(1,2,2)

scatter(YTest,YPredLM,50,'filled')

Add your label

[Kamal @Kamall282]

Let's build a Neural network from scratch

Let's use Neural Network to predict Fuel consumption of a vehicle engine.

Let's use these input parameters

Engine rpm, vehicle speed, temperature, throttle, air intake rate, vehicle mass, tyre pressure, road slope.

Output: fuel rate

[Kamal @Kamall282]

mdl = fitlm(XTrain,YTrain);

YPredLM = predict(mdl,XTest);

RMSE_LM = sqrt(mean((YTest-YPredLM).^2));

There's a need to visualize the neural net Vs Linear regression model

[Kamal @Kamall282]

Which on a thorough test, the model might drop to 40% 💀💀😂

Experienced ML guys don't trust training accuracy, but the performance of unseen data.

[Kamal @Kamall282]

Partitioning of data in ML algorithm.

Before fitting any ML model, ask yourself:
“How do I know my model will work on new data?”

This is exactly why MATLAB has cvpartition() command

Think of your dataset as 100 students.
If you teach all 100 students and then test the same 100

[Kamal @Kamall282]

mdl = fitlm(Xtrain,Ytrain);

We then perform a real test

Ypred = predict(mdl, Xtest);

This is data the model has never seen before, the performance here is what actually matters.

Without cvpartition(), you might get 99% accuracy and you'll think you've built a real model