https://github.com/u66u/lotus

ML framework for simple neural networks with optimal defaults
https://github.com/u66u/lotus

ai artificial-intelligence cpp learning-resources machine-learning ml neural-network neural-networks open-source pytorch tensorflow tf torch transformers tutorial

Last synced: 27 days ago
JSON representation

ML framework for simple neural networks with optimal defaults

• Host: GitHub
• URL: https://github.com/u66u/lotus
• Owner: u66u
• Created: 2023-09-20T16:06:32.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-10-10T17:01:47.000Z (about 1 year ago)
• Last Synced: 2023-11-09T15:32:16.348Z (about 1 year ago)
• Topics: ai, artificial-intelligence, cpp, learning-resources, machine-learning, ml, neural-network, neural-networks, open-source, pytorch, tensorflow, tf, torch, transformers, tutorial
• Language: C++
• Homepage:
• Size: 62.5 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
set different ranges for different weght

bias inits, set default ranges for init functions

test memory

check iterators in layer

Implement differen weight/bias init for xavier:

The original paper by Glorot and Bengio suggests a variance of 2/(n_in + n_out) where n_in is the number of inputs and n_out is the number of outputs of the neuron 365datascience.com.

Some sources, such as the deeplearning.ai notes, suggest a variance of 1/n_in deeplearning.ai.

Some other sources, such as machinelearningmastery.com, suggest a uniform distribution in the range -(sqrt(6)/sqrt(n_in + n_out)) to sqrt(6)/sqrt(n_in + n_out) machinelearningmastery.com.

print statements are triggered even if the ranges are defined for init methods