PHi-C v2.2.0

⚡ Performance Improvements

Dramatically faster optimization
The optimization algorithm has been completely overhauled:

• Replaced fixed-step gradient descent with a backtracking line search, enabling adaptive step sizes and robust convergence
• Replaced the eigh-based pseudoinverse with a Cholesky decomposition, significantly reducing computation time per step
• New options: --backtracking-factor and --gradient-degree; --learning-rate has been removed

Vectorized MSD and loss tangent computation

• msd and losstangent now use fully vectorized matrix operations instead of per-bead for loops, resulting in orders-of-magnitude speedup for large matrices
• Results are stored in a single compressed .npz file instead of per-bead .txt files

---

🔄 Subcommand Changes

Consolidated rheology subcommands

• Removed: rheology, plot-tangent, plot-compliance, plot-modulus
• Added: losstangent, plot-losstangent

The new losstangent subcommand focuses on the loss tangent (tan δ) spectrum, which is the most informative quantity for characterizing local chromatin dynamics.

---

📁 Output Format Changes

All data output files have been changed from plain text (.txt) to NumPy compressed format (.npz):

• C_normalized.npz, P_normalized.npz
• K_optimized.npz, C_optimized.npz
• MSD_matrix.npz, losstangent_matrix.npz

---

✨ New Features

• dynamics and sampling now additionally output .dcd trajectory files (via MDAnalysis)
• plot-optimization now outputs Eta.svg showing the learning rate trajectory
• Progress bars added to dynamics and sampling via tqdm

---

📦 New Dependencies

• MDAnalysis
• tqdm

---