ML2018FALL

ML is so difficult

HW1 - PM2.5 prediction

Using numpy to construct a basic linear model predicting PM2.5.

---

HW2 - Default Payment Prediction

Using numpy to contruct a simple classifier predicting credit card payment.

---

HW3 - Image Sentiment Classification

Using Keras to build a CNN model to classify the face expression of image.

---

HW3 bonus - Hyperparameter Tuning (PyTorch)

Tune the hyperparameters to achieve higher performance on LSUN dataset with ResNet50 and MobileNet.

---

HW4 - Malicious Comments Identification

Using Keras to build a RNN model to classify if the comments are malicious.

---

Final Project - Small Data Training for Medical Images

Using pytorch to build a CNN model that classify chest diseases with DenseNet201 and Auto-Encoder. We achieved a AUROC=0.79998 score using only small part of NIH dataset. Colaboration with

NTUEE B05901009 高瑋聰

NTUEE B05901034 劉奎元

View our code from

https://github.com/sky1456723/ML2018FALL