Aleph - 0.36.0

🚀 Highlights

---

✨ New Features

• Add language server support for formatting Aleph code files (beta).

• Added logarithmic binning accumulators to statistics support. These can be used to estimate auto correlation times and can be constructor by calling accumulator("logarithmic"). Includes parameters base and initial_depth to set the bin size of each level of the logarithmic accumulation and the initial depth of the levels.

• MPS and MPO arithmetic operations now support integer, real and complex scalar types.

  • In-place product and division with scalar: mps *= scalar and mps /= scalar, mpo *= scalar and mpo /= scalar.
  • Symmetric scalar multiplication: scalar * mps, scalar * mpo, mps * scalar and mpo* scalar.
  • Division by scalar: mps / scalar and mpo / scalar.
  • Tensors now support in-place element-wise scalar arithmetic operations: +=, -=, *=, /=.

• Integrated aliased discrete distribution into Aleph which offers O(1) sampling speed as opposed to O(N) from a standard discrete distribution. Can be built using random_distribution("aliased_discrete", ["probabilities" : [...]]).

• Universal operator identity and zero factories: Identity() and Zero().

• More fermionic operator transformations have been added:

  • anticommuted : anticommutes two fermionic operators $A,B$ and returns $\{ A , B \} - BA$ while respecting the canonical commutation relations of $f,f^{\dag}$. Supported for Create, Destroy, and Number operators.
  • site_canonicalized : orders a a fermionic operator into site canonical order $\prod_k (f^{\dag}_k)^{m_k} (f_k)^{n_k}$ with $m_k,n_k \in \{ 0,1 \}$.
  • canonicalize : orders a site-canonical product into complete canonical order $\prod_k f^{\dag}_{k} \prod_l f_l$.

• Added a size() method to OperatorProducts to enable easier for looping over terms.

---

🐛 Fixes

• Fixed an error that prevented creating or writing HDF5 files in subfolders.

---

⚙️ Internal / Technical Changes

• Aleph now uses the Xoshiro256 algorithm as the default random number generator as opposed to Mersenne Twister. Xoshiro offers generally superior performance in terms of both time and space requirements.

---

⚠️ Breaking Changes

• Removed ID and FermiId in favor of the unified Identity() operator factory. Use Identity() for universal identity operators across all operator types.

---