Topics
Thematic anchors for the Shiplog. Each topic groups related papers, talks, and notes; the Shiplog visualizes how they connect.
adjacent explorations
- curiosity Adjacent explorations
Side-track exploration outside the random-matrix / ML core — VR, haptic interfaces, projection mapping geometry.
- theme VR, haptics and projection mapping
VR exploration, haptic interfaces, and projector-placement geometry — the concrete strands inside the adjacent-explorations bucket.
free probability and random matrices
- theme Self-attention spectra
Spectral and signal-propagation analysis of transformer self-attention layers; Gaussian-equivalence and deviation from Marchenko-Pastur.
- curiosity Free probability and random matrices
Operator-algebraic and combinatorial tools for the spectral behavior of large random matrices.
- concept Random matrix spectra
Asymptotic eigenvalue distributions of structured random matrices.
- concept Free group representation
Approximate representations generated by random orthogonal matrices.
- concept Asymptotic freeness
When independent random matrices become free in the large-dimensional limit, justifying free-probability propagation through DNN layers.
- theme Quantum probability and easy quantum groups
Boolean / free / classical independence and the easy-quantum-group symmetry underlying de Finetti-type theorems.
- theme Random matrix parameter estimation
Recovering random-matrix model parameters from spectral statistics — identifiability, Cauchy-noise losses, and free deterministic equivalents.
machine learning
- curiosity Machine learning
DNN theory through random-matrix lenses — free random projection, dynamical isometry, meta-RL, training dynamics.
- theme Dynamical isometry
Conditions under which signal propagation in deep networks preserves norms and gradients; spectral analysis of layerwise Jacobians and Fisher information.
- method Orthogonal initialization
Initializing weight matrices as random orthogonal matrices to preserve singular values.
- theme DNN architectures as random-matrix systems
Reading deep architectures (MLP-Mixer, attention, sparse MLPs) through random-matrix and Kronecker-structure lenses to expose implicit regularization.
- method Free Random Projection
Random representation-based projection method for in-context and meta-reinforcement learning.
- theme Meta reinforcement learning
Learning algorithms that adapt to new tasks from limited interaction.
- theme Reinforcement learning
Sequential decision-making under uncertainty — the umbrella over meta-RL adaptive learning and the VR-scene exploration policies in the adjacent thread.
- theme Interpretability and training dynamics
Layer-wise interpretability via identity initialization, implicit bias of gradient regularization, and selective forgetting / unlearning.