Todayβs ML progress
Went from logits β probabilities β prediction labels, improved my classification model, and tested how well it learns.
The big takeaway: non-linearity is what lets neural networks learn more complex patterns.
#100DaysOfCode
The hardest part of learning ML isn't writing the code.
It's building enough intuition to know what the code is actually doing.
That's what I'm trying to improve now. π§ π€
Today I built my first classification neural network with PyTorch! π€
Learned about classification data, model architecture, loss functions, optimizers, and evaluation.
Slowly moving from regression to deeper ML concepts. π
#PyTorch#100DaysOfCode
Today I went deeper into PyTorch training loops. π€
Learned how to:
β’ Build training & testing loops
β’ Run models epoch by epoch
β’ Understand what happens at each step
β’ Save trained PyTorch models
Starting to understand what happens under the hood. π
#100DaysOfCode
Today I built my first Linear Regression model with PyTorch. π€
Learned how to create a dataset, split train/test data, visualize it, make predictions, and understand whatβs needed to train a model.
Slowly putting the pieces together. π
#PyTorch#100DaysOfCode
Day by day, getting deeper into tensors π€
Today I learned reshaping, stacking, squeezing, unsqueezing, permuting, indexing, and working with PyTorch & NumPy.
Small progress every day. π
#100DaysOfCode#PyTorch
Todayβs ML learning π€
Scalar β Vector β Matrix
Mean, Max, Min + basic arithmetic operations.
Small steps, but making progress. π
#100DaysOfCode#ML#AI