Neural Networks with CasADi

csnn is a package for creating symbolic neural networks in CasADi in a PyTorch-like API style.

---

Introduction

The package allows the creation of neural networks with the symbolic language offered by CasADi. This is done in a similar way to PyTorch. For example, the following code allows us to create an MLP with a hidden layer:

import casadi as cs
from csnn import set_sym_type, Linear, Sequential, ReLU

set_sym_type("SX")  # can set either MX or SX

net = Sequential[cs.SX]((
    Linear(4, 32),
    ReLU(),
    Linear(32, 1),
    ReLU()
))

batch = 2
input = cs.SX.sym("in", batch, 4)
output = net(input)
assert output.shape == (batch, 1)

---

Implemented Modules

So far, the following modules that are available in PyTorch have been implemented:

• Containers
  • Module
  • Sequential
• Activation functions
  • GELU
  • SELU
  • LeakyReLU
  • ReLU
  • Sigmoid
  • Softplus
  • Tanh
• Linear layers
  • Linear
• Recurrent layers
  • RNNCell
  • RNN
• Dropout layers
  • Dropout
  • Dropout1d

Additionally, the library provides the implementation for the following convex neural networks (see csnn.convex):

• FicNN
• PwqNN
• PsdNN

---

Installation

To install the package, run

pip install csnn

csnn has the following dependencies

• CasADi

For playing around with the source code instead, run

git clone https://github.com/FilippoAiraldi/casadi-neural-nets.git

---

License

The repository is provided under the MIT License. See the LICENSE file included with this repository.

---

Author

Filippo Airaldi, PhD Candidate [f.airaldi@tudelft.nl | filippoairaldi@gmail.com]

Delft Center for Systems and Control in Delft University of Technology

Copyright (c) 2023 Filippo Airaldi.

Copyright notice: Technische Universiteit Delft hereby disclaims all copyright interest in the program “csnn” (Nueral Networks with CasADi) written by the Author(s). Prof. Dr. Ir. Fred van Keulen, Dean of 3mE.