Machine Learning Blogs_6

-

Hi there! So this one is my sixth post of Machine learning concepts learning in a creative way. I hope so you are enjoying my blogs and I am also open to receive your opinions. 

And again if you want to correct something about my post , feel free to comment out.

Ready to learn todays important topics? We can actually refer them as the gang of students who helps to do the task and produce great output which is actually need as the output!

The legends are known as : " Activation Functions" Now its time to learn it in a creative way that you would never forget even if you want too 😉

Activation Functions: The Rockstars of Neural Nets ðŸŽ¸

Why does a neural net need these? Because without them, it’s just linear algebra—boring!

The Band Members:

Sigmoid (The Vintage Rocker)

  • Sound: Smooth 70s vibes (S-shaped curve)

  • Good For: Probability (0 to 1)

  • Downside: "Dying gradient" problem (gets lazy in deep networks)

ReLU (The Punk Rocker) ⚡

  • Sound: "If it’s negative, ZERO! If positive, ROCK ON!"

  • Good For: Deep learning (fast, avoids vanishing gradients)

  • Downside: "Dead neurons" (some stop working forever)

Leaky ReLU (The Emo Rocker)

  • Sound: "Even negatives get a tiny chance..."

  • Fix: Prevents dead neurons by allowing small negatives

Tanh (The Jazz Musician)

  • Sound: Smooth but edgy (-1 to 1)

  • Good For: Hidden layers

  • Downside: Still suffers vanishing gradients

Softmax (The Choir Director)

  • Sound: "Everyone gets a probability, and they sum to 1!"

  • Good For: Multi-class outputs (e.g., classifying cats/dogs/birds)

Why ReLU is the Lead Singer

  1. Fast to compute (no exponentials like sigmoid)

  2. Sparsity (some neurons "turn off"—efficient!)

  3. Solves vanishing gradients in deep nets

(But sometimes needs backup singers—like Leaky ReLU—to cover its flaws!)

Final Jam Session:

  • Gradient descent = How the net learns (gym routine)

  • Activation functions = How it makes decisions (rockstar flair)

Enjoyed this post? Next be Ready to get to know more and something interesting like Backpropagation and Dropout .
Will meet you next time! Unitl then HAPPY LEARNING!!

Comments

Post a Comment

Popular posts from this blog

Machine Learning Blogs_2

-
 Core Terms of Machine Learning Well hello there and welcome back! As I promised in previous blog , I will be describing the core topics or terms you should know how to define and be smarter than you overconfident nerd friend 😉 Model : " The part of Machine Learning that learns and makes predictions is called Model " easy right? now you know what model is! But examples? here's the example for ya'll : Neural Networks ,Random forests, etc . Data Mining : " Digging into large amounts of data to discover hidden patterns . " Transfer Learning: " Transferring knowledge from one task to another is called transfer learning "  Labels: "The training set you feed to the algorithm includes desired solutions." Supervised Learning: " The system tries to learn with the teacher." Not so technical definition ,but I am also not going into technical stuff soo early. Unsupervised learning: " The system tries to learn without teacher." an...
Read more
-
Well hello there! I am Soham and fresher Graduate in the field of AI and Machine Learning. But just as you I was a student to who was just as confused as you are when I had started to learn AI. But see here I am writing something for you about AI and Machine Learning to explain you just how amazing building an AI is ! And No I will not teach you how to do it ,because there are tons of people out there on YouTube who are doing it for free. I am just here who will try to clear your concepts and also in a creative way that you would never forget it 😉. I will also share some technicalities so there wont be half cooked knowledge of theory. Just sharing it all from a friend to another friend who is eager to know something. GoodLuck ! What is Machine Learning? Many People might think its just bunch of algorithms used for training Machine to do certain task, right? But directly moving to algorithms and training? is it right for complete beginner to feed so much complex...
Read more