@beatricegideonblessed: "We do not pray because of troubles; sometimes you go before God praying for transformation #PrayerAndFasting #ApostleJohnKimaniWilliam

blessed
blessed
Open In TikTok:
Region: KE
Monday 05 October 2026 07:18:18 GMT
4761
320
2
38

Music

Download

Comments

user4559779439355
user4559779439355 :
v true... well put
2026-10-05 16:09:26
0
_peninah1
peninah❤️ :
very true. N well noted.
2026-10-06 07:13:44
0
To see more videos from user @beatricegideonblessed, please go to the Tikwm homepage.

Other Videos

➡️ Machine Learning from Scratch part 25: a real deep network Our code drew 400 points on two spirals wound into each other, 200 blue and 200 orange. The best possible straight line gets only 267 of them right, so a single neuron has no chance. The network has 3 inputs (x, y and a constant 1 that lets every line shift), hidden layers of 16, 16 and 3 neurons, and 1 output: 355 weights, random at the start. Each neuron weighs all the numbers from the layer before, adds them up and puts the sum through tanh. In each of 10,000 training steps the error flows back through every layer (backpropagation) and every weight takes a small step downhill. The network starts at 200 of 400, stumbles to 188 around step 857, gets all 400 right at step 1,613 and stays there from step 1,631 on. Following one new blue point at (0.611, 0.097): layer 1 draws 16 straight lines and checks which side the point is on, layer 2 combines those answers into bends and curves, and layer 3 squeezes everything into three numbers. In that space one flat plane separates all 400 points. The output reads 0.96, close to +1, so blue. An orange point ends at -1.00. The real test: 200 new points it has never seen, 200 of 200 right. Is that really deep? Deep learning means models consisting of multiple layers of neurons, so three hidden layers make a small deep network. We also gave the same 35 neurons one wide layer: on 20 other spiral sets it learned all 400 points only 8 times, the deep network all 20 times. #machinelearning #deeplearning #neuralnetwork #ai #python
➡️ Machine Learning from Scratch part 25: a real deep network Our code drew 400 points on two spirals wound into each other, 200 blue and 200 orange. The best possible straight line gets only 267 of them right, so a single neuron has no chance. The network has 3 inputs (x, y and a constant 1 that lets every line shift), hidden layers of 16, 16 and 3 neurons, and 1 output: 355 weights, random at the start. Each neuron weighs all the numbers from the layer before, adds them up and puts the sum through tanh. In each of 10,000 training steps the error flows back through every layer (backpropagation) and every weight takes a small step downhill. The network starts at 200 of 400, stumbles to 188 around step 857, gets all 400 right at step 1,613 and stays there from step 1,631 on. Following one new blue point at (0.611, 0.097): layer 1 draws 16 straight lines and checks which side the point is on, layer 2 combines those answers into bends and curves, and layer 3 squeezes everything into three numbers. In that space one flat plane separates all 400 points. The output reads 0.96, close to +1, so blue. An orange point ends at -1.00. The real test: 200 new points it has never seen, 200 of 200 right. Is that really deep? Deep learning means models consisting of multiple layers of neurons, so three hidden layers make a small deep network. We also gave the same 35 neurons one wide layer: on 20 other spiral sets it learned all 400 points only 8 times, the deep network all 20 times. #machinelearning #deeplearning #neuralnetwork #ai #python

About