/AN2DL

Challenges of the Artificial Neural Networks and Deep Learning at Politecnico di Milano

Primary LanguageJupyter Notebook

AN2DL

Hot Neurons Group 🔥

Image Classification Challenge 🍃

The objective of this challenge was to correctly classify images of leafs based on the species of the plant they belong.
We developed three different solutions for this problem :

  • Convolutional Neural Network from scratch
  • Transfer Learning approach with Adam optimiser
  • Transfer Learning approach with SGD optimiser

Our best score was 0.9434 obtain with SGD and Adam.

Time-Series Forecasting Challenge ⏱

The objective of this challenge was to forecast a composition of seven signals.
We developed three different solutions for this problem :

  • LSTM Network with Direct Forecating
  • LSTM Network with Autoregressive Forecasting
  • Transformer

Our best MSE was 3.78 with the Autoregressive LSTM Network.

Grade 🎓

Total grade of both challenges : 10/11.

Devs 💻