TildAlice – Tech Blog for Developers
Python, AI, Deep Learning, Data Analysis, Computer Vision, Reinforcement Learning โ tips and tutorials
Latest Posts
-
RUL Prediction: Classical ML vs LSTM Training Guide
Random Forest beat LSTM by 12% MAE in offline testsโbut LSTMs won in streaming inference. Here's the benchmark data and when to use each.
-
Feature Engineering for Predictive Maintenance: Health Indicators
Extract FFT peaks, RMS trends, and thermal gradients from vibration data for RUL predictionโfeatures that beat raw signals by 40% accuracy.
-
CBM from Scratch: Sensor Data Collection and Preprocessing
Fix the 3 timestamp sync errors that ruin CBM predictions. Trigger-based logging pipeline design with sub-second precision on edge devices.
-
ONNX Runtime๊ณผ TensorRT๋ก ์ปดํจํฐ๋น์ ๋ชจ๋ธ ๊ฒฝ๋ํํ๊ธฐ: Edge AI ๋ฐฐํฌ ์ค์ ๊ฐ์ด๋
Speed up YOLO inference 8x on Jetson with TensorRT INT8 quantization. ONNX conversion pitfalls and FP16 vs INT8 accuracy tradeoffs included.
-
์ฝ๋ฉํ ์คํธ ์๋ฒฝ ๋๋น ๊ฐ์ด๋: ์๊ณ ๋ฆฌ์ฆ ์ ํ๋ถํฐ ๋ฌธ์ ํ์ด ์ ๋ต๊น์ง
Solve Two Sum in O(n) with hash maps instead of O(nยฒ) brute force. Compare 3 approaches with time complexity proof and LeetCode patterns.
-
Model-Free vs Model-Based RL: ์ํ ํจ์จ์ฑ ๋น๊ต ๊ฐ์ด๋
Model-Free vs Model-Based RL: which needs fewer samples? Atari/MuJoCo benchmarks reveal the trade-offs you need to know before coding.
-
Exploration vs Exploitation: ฮต-greedy๋ถํฐ UCB, Thompson๊น์ง
Compare ฮต-greedy, UCB, Thompson Sampling, and RND for sparse rewards. See which exploration strategy actually works in hard environments.
-
VAE๋ก ์ ์ ๋ฐ์ดํฐ๋ง์ผ๋ก ํฐ๋น ๋ฒ ์ด๋ง ๊ณ ์ฅ ๊ฐ์งํ๊ธฐ: ๋ฌด๋ผ๋ฒจ ์ด์ ํ์ง ์ค์ ๊ฐ์ด๋
Detect turbine bearing failures using only normal data. VAE reconstruction error approach with Python code โ no labels required.