In this page, you will find all the previous reading group session.

2022 Q2: 30/05/22

• Tiny ML : Machine Learning for embedded systems by Yanis Slides here

• RPi cluster & Deep Learning by Fatou Kiné Slides here

• Nvidia Jetson Nano by Nicolas Slides here

2022 Q2: 28/03/22

• Sparse matrix multiplication with pytorch and Cuda by Nicolas and Yanis (14h)

• Inference using tflite with larq by Fatou (14h30)
  Slides here

• Social computing : NLP and community detection by Matthieu and Paul (15h) Slides here

• Greedy decentralized optimization for deep learning by Simon (16h00)

• Convergence of MH algorithm for deep learning by Jordy (16h30)

2022 Q1: 24/01/22

• Binarization of Neural Networks by Fatou Kiné Sow & Matthieu François

Slides here

[1] « XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks », M. Rastegari, V. Ordonez, J. Redmon, A. Farhadi. See the full article here.

[2] « XNOR-Net++: Improved binary neural networks » , A. Bulat, G. Tzimiropoulos. See the full article here.

[3] « BinaryConnect: Training Deep Neural Networks with binary weights during propagations », M. Courbariaux, Y. Bengio, J-P. David. See the full article here.

[4] « Binarized Neural Networks: Training Neural Networks with Weights and Activations Constrained to +1 or −1 », M. Courbariaux, I. Hubara, D. Soudry, R. El-Yaniv, Y. Bengio. See the full article here.

• Early Exit and Device to Cloud by Simon Lebeaud

Slides here

For more details, see :

[1] « Multi-scale dense networkds for resource efficient image classification » , G. Huang, D. Chen,T. Li, F. Wu L. van der Maaten, K. Weinberger. See the full article here.

[2] « SPINN : Synergistic Progressive Inference of Neural Networks over device and cloud », S. Laskaridis, S. I. Venieris, M. Almeida, I. Leontiadis, N. D. Lane. See the full article here.

• Pruning by Nicolas Tirel & Yanis Chaigneau

See the Jupyter-notebook here.

[1] « Skeletonization: A Technique for Trimming the Fat from a Network via Relevance Assessment », M. C. Mozer et P. Smolensky, 1988. See the complete article.

[2] « SNIP: Single-shot Network Pruning based on Connection Sensitivity », N. Lee, T. Ajanthan, P. H. S. Torr. See the full article here.

[3] « Progressive Skeletonization: Trimming more fat from a network at initialization », P. de Jorge, A. Sanyal, H. S. Behl, P. H. S. Torr, G. Rogez, et P. K. Dokania, March 2021. See the full article here.

• Pruning with budget and Regularization : Jordy Palafox [1] « Deep Rewiring: Training very sparse deep networks », G. Bellec, D. Kappel, W. Maass, R. Legenstein. See the full article here.

[2] « Statistical guarantees for regularized neural networks », M. Taheri, F. Xie, J. Lederer. See the full article here.

2021 Q4: 29/10/2021

• Metrical Task System, Online Learning and Power Management by Matthieu François :

[1] « On-line Learning and the Metrical Task System Problem », A. Blum, C. Burch. See the full article here.

[2] « Online Strategies for Dynamic Power Management in Systems with Multiple Power-Saving States », S. Irani, S. Shukla, R. Gupta. See the full article.

• Metrical Task System and K-server problem by Jordy Palafox :

[1] « An Optimal On-Line Algorithm for Metrical Task System », A. Borodin, N. Linial, M.E. Saks. See the full article here.

[2] « Competitive Algorithms for Server Problems », M.S. Manasse, L.A. McGeoch, D.D. Sleator. Journal of algrithms 11, 208-230 (1990). See the full article here.

Contact

You want to join the team ? Feel free to contact us if you want to contribute: contact Paul or Sébastien