Pith. sign in

REVIEW 6 cited by

On weight initialization in deep neural networks

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1704.08863 v2 pith:Q6EIRBG4 submitted 2017-04-28 cs.LG

On weight initialization in deep neural networks

classification cs.LG
keywords initializationweightneuralactivationactivationsderivefunctionsinitializations
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

A proper initialization of the weights in a neural network is critical to its convergence. Current insights into weight initialization come primarily from linear activation functions. In this paper, I develop a theory for weight initializations with non-linear activations. First, I derive a general weight initialization strategy for any neural network using activation functions differentiable at 0. Next, I derive the weight initialization strategy for the Rectified Linear Unit (RELU), and provide theoretical insights into why the Xavier initialization is a poor choice with RELU activations. My analysis provides a clear demonstration of the role of non-linearities in determining the proper weight initializations.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. $\text{DT}^2$: Decision-Targeted Digital Twins

    cs.LG 2026-06 unverdicted novelty 7.0

    DT² trains digital twins to preserve pairwise policy rankings from fitted Q-evaluation on offline data rather than minimizing one-step transition errors, improving policy ranking and reducing decision regret.

  2. Efficient Unlearning through Maximizing Relearning Convergence Delay

    cs.LG 2026-04 unverdicted novelty 7.0

    The Influence Eliminating Unlearning framework maximizes relearning convergence delay via weight decay and noise injection to remove the influence of a forgetting set while preserving accuracy on retained data.

  3. An Extensible and Lightweight Unified Architecture for Demosaicing Pixel-bin Image Sensors

    cs.CV 2026-06 unverdicted novelty 6.0

    Proposes a modular unified lightweight neural architecture for demosaicing multiple pixel-bin CFAs together with a learning-free CFA identification module.

  4. Critical evaluation of PINN for FWD inverse analysis and differentiable FEM as an alternative

    cs.CE 2026-06 unverdicted novelty 6.0

    Standard PINNs fail to recover layer moduli in multilayer pavement systems due to domain discontinuities; XPINN improves but remains sensitive to weighting and noise; DiffFEM yields superior accuracy, stability, and e...

  5. Trainability of IQP Quantum Circuit Born Machines Under Gaussian Initialization

    quant-ph 2026-06 unverdicted novelty 5.0

    Derives lower bound on gradient variance and probabilistic concentration bounds for Gaussian-initialized IQP QCBMs trained via MMD loss.

  6. Embedding Non-Distortive Cancelable Face Template Generation

    cs.CV 2024-02 unverdicted novelty 3.0

    Presents a non-distortive cancelable face template method via targeted image distortion that maintains identity signals for neural embedding models on MNIST and LFW data.