This post verifies the core writing surface for a deep learning research blog: mathematical notation, code blocks, external links, generated math artifacts, and a JavaScript simulation.

Inline math works with passthrough delimiters: ∇θL(θ) \nabla_\theta \mathcal{L}(\theta) .

Display math works as well:

L(θ)=−∑i=1nyilog⁡pθ(yi∣xi) \mathcal{L}(\theta) = - \sum_{i=1}^{n} y_i \log p_\theta(y_i \mid x_i)

Code blocks use Hugo syntax highlighting and PaperMod copy controls:

import torch

def cosine_similarity(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:
    x = torch.nn.functional.normalize(x, dim=-1)
    y = torch.nn.functional.normalize(y, dim=-1)
    return x @ y.T

External links such as Hugo are rendered with safer external-link attributes.

Static artifacts can live next to the post.

Lorenz projection generated as a static artifact

Loading simulation...