class SrinjoyRoy:
def __init__(self):
self.name = "Srinjoy Roy"
self.username = "srinjoy356"
self.pronouns = "he/him"
self.location = "India 🇮🇳"
self.interests = ["Data Science", "Machine Learning", "Statistics", "Neural Networks"]
self.learning = ["ML Algorithms", "Deep Learning", "Real-world Projects"]
self.collaborate = ["Python", "ML", "Software Development"]
self.philosophy = "I don't learn traditionally — I learn by doing."
def loss_fn(self, knowledge: float) -> float:
"""Minimize ignorance. Maximize understanding."""
return 1 / (1 + knowledge) # → 0 as knowledge → ∞
def gradient_step(self):
self.knowledge += self.curiosity * self.learning_rate
def __repr__(self):
return f"∂(impact)/∂(effort) > 0 → Still optimizing 🚀"
me = SrinjoyRoy()Where
$x_i$ = experience,$w_i$ = curiosity,$b$ = stubbornness to give up,$\sigma$ = the real world
- 🔭 Currently: Training models, debugging loss curves, building projects
- 📊 Domain: ML · GenAI · Computer Vision · NLP · MLOps · Statistics
- 🌐 Portfolio: srinjoy-roy.vercel.app
- 🌱 Gradient Descent: Always descending toward deeper understanding
- 💡 Philosophy:
loss → 0only when you stop following the textbook - 🤝 Open to: Collaborations in Python, ML, or any software project worth building
- ⚡ Fun fact: My learning rate is adaptive — I skip the boring parts
Epoch 1 : "What is a neural network?" Loss: ████████████ 0.98
Epoch 10 : "I can train a model!" Loss: ████████░░░░ 0.72
Epoch 50 : "Hyperparameter tuning hurts..." Loss: ████░░░░░░░░ 0.41
Epoch ∞ : Still running... Loss: Converging → 0
$\mathcal{L}(\theta) = -\frac{1}{m}\sum_{i=1}^{m}\left[y^{(i)}\log\hat{y}^{(i)} + (1-y^{(i)})\log(1-\hat{y}^{(i)})\right]$
"All models are wrong, but some are useful." — George Box