Generalized CP Optimization (pyttb.gcp_opt)

Note

The gcp_opt function defined in gcp_opt.py has been promoted to the pyttb namespace.

pyttb.gcp_opt(data: ttb.tensor | ttb.sptensor, rank: int, objective: Objectives | tuple[function_type, function_type, float], optimizer: StochasticSolver | LBFGSB, init: Literal['random'] | ttb.ktensor | Sequence[np.ndarray] = 'random', mask: ttb.tensor | np.ndarray | None = None, sampler: GCPSampler | None = None, printitn: int = 1) → tuple[ttb.ktensor, ttb.ktensor, dict]

Fits Generalized CP decomposition with user-specified function.

Parameters:

• data – Tensor to decompose.

• rank – Rank of desired CP decomposition.

• objective – Objective function to minimize for the CP decomposition. Either a pre-defined objective or a tuple of function_handle, gradient_handle, and lower_bound.

• optimizer – Optimizer class for solving the decompistion problem defined.

• init – Initial solution to the problem.

• mask – A binary mask to note missing rather than sparse data. (Only valid for dense, LBFGSB solves)

• sampler – Class that defined sampling strategy for stochastic solves.

• printitn – Controls verbosity of printing throughout the solve

Returns:

Solution, Initial Guess, Dictionary of meta data

• pyttb.gcp.fg
  • evaluate()
• pyttb.gcp.fg_est
  • estimate()
  • estimate_helper()
• pyttb.gcp.fg_setup
  • setup()
  • valid_nonneg()
  • valid_binary()
  • valid_natural()
• pyttb.gcp.handles
  • Objectives
    • Objectives.GAUSSIAN
    • Objectives.BERNOULLI_ODDS
    • Objectives.BERNOULLI_LOGIT
    • Objectives.POISSON
    • Objectives.POISSON_LOG
    • Objectives.RAYLEIGH
    • Objectives.GAMMA
    • Objectives.HUBER
    • Objectives.NEGATIVE_BINOMIAL
    • Objectives.BETA
    • Objectives.ZT_POISSON
  • gaussian()
  • gaussian_grad()
  • bernoulli_odds()
  • bernoulli_odds_grad()
  • bernoulli_logit()
  • bernoulli_logit_grad()
  • poisson()
  • poisson_grad()
  • poisson_log()
  • poisson_log_grad()
  • rayleigh()
  • rayleigh_grad()
  • gamma()
  • gamma_grad()
  • huber()
  • huber_grad()
  • negative_binomial()
  • negative_binomial_grad()
  • beta()
  • beta_grad()
  • ztp()
  • ztp_grad()
• pyttb.gcp.optimizers
  • StochasticSolver
    • StochasticSolver.__init__()
    • StochasticSolver.update_step()
    • StochasticSolver.set_failed_epoch()
    • StochasticSolver.solve()
  • SGD
    • SGD.update_step()
    • SGD.set_failed_epoch()
  • Adam
    • Adam.__init__()
    • Adam.set_failed_epoch()
    • Adam.update_step()
  • Adagrad
    • Adagrad.__init__()
    • Adagrad.set_failed_epoch()
    • Adagrad.update_step()
  • LBFGSB
    • LBFGSB.__init__()
    • LBFGSB.solve()
    • LBFGSB.Monitor
      • LBFGSB.Monitor.__init__()
      • LBFGSB.Monitor.callback
• pyttb.gcp.samplers
  • StratifiedCount
    • StratifiedCount.num_zeros
    • StratifiedCount.num_nonzeros
    • StratifiedCount.__init__()
  • Samplers
    • Samplers.UNIFORM
    • Samplers.SEMISTRATIFIED
    • Samplers.STRATIFIED
  • GCPSampler
    • GCPSampler.__init__()
    • GCPSampler.function_sample()
    • GCPSampler.gradient_sample()
    • GCPSampler.crng
  • nonzeros()
  • zeros()
  • uniform()
  • semistrat()
  • stratified()