o .wi=@sddlmZmZmZddlZddlmZddlmZddlm Z ddl m Z ddl m Z mZmZddlmZGd d d eZGd d d eZd ededeeeffddZd edededefddZdededefddZd edededefddZdededefddZdGd ededefdd Zd edeeeeffd!d"Zdeded#edeeeeeeeeeeeeeeeef fd$d%Zdeded&ed'ed(ed)ed*eed+eed#edefd,d-Zd.edefd/d0Z d(ed)ed*eed1eed2eed+eed3eed4eed#ed5eedefd6d7Z!   8dHdeded9eeed:eeed;edeeeeeff dd?Z# @  AdJdeded#edBdCed5eedDdeeeeefff dEdFZ$dS)K)ListOptionalUnionN)Tensor)Literal) _check_data_shape_to_num_outputs)_check_same_shape) _bincount_cumsum dim_zero_cat)EnumStrc@.eZdZdZdZdZdZedefddZ dS) _MetricVariantzEnumerate for metric variants.abcreturncCdS)Nvariantrrrg/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/torchmetrics/functional/regression/kendall.py_name!z_MetricVariant._nameN) __name__ __module__ __qualname____doc__ABC staticmethodstrrrrrrrrc@r ) _TestAlternativez'Enumerate for test alternative options. two-sidedlessgreaterrcCr)N alternativerrrrrr-rz_TestAlternative._nameN) rrrr TWO_SIDEDLESSGREATERr r!rrrrrr#&r"r#xyrcCsXt|}|j|j}}|\}}t|jdD] }||||||<q|j|jfS)zBSort sequences in an ascent order according to the sequence ``x``.r)torchcloneTsortrangeshape)r+r,permirrr_sort_on_first_sequence2s   r5r4cCs@t||||ddk||||ddkddS)z>Count a total number of concordant pairs in a single sequence.Nr)r- logical_andsum unsqueezer+r,r4rrr_concordant_element_sum=s@r;predstargetc*tfddtjdDdS)zDz+_count_concordant_pairs..rr-catr1r2r8rBrrBr_count_concordant_pairsB*rGc Csxtt||||ddk||||ddkt||||ddk||||ddkddS)z?Count a total number of discordant pairs in a single sequences.r6Nr)r- logical_orr7r8r9r:rrr_discordant_element_sumGs22rJcr>)z.rrErBrrBr_count_discordant_pairsSrHrKFr0cCs`|r |jddj}tjd|jdtj|jd}ttj||dd|ddk gddddS)z&Convert a sequence to the rank tensor.r)dimr6dtypedeviceN) r0valuesr-zerosr2int32rOr rFint)r+r0_onesrrr_convert_sequence_to_dense_rankXs2rVcCstj|jd|j|jd}tj|jd|j|jd}tj|jd|j|jd}t|jdD]<}t|dd|f}||dk}||dd||<||d|d||<||dd|d||<q.|||fS)zWGet a total number of ties and staistics for p-value calculation for a given sequence.r6rMN?)r-rRr2rNrOr1r r8)r+tiesties_p1ties_p2rLn_tiesrrr _get_tiesas " r^rc Cst||\}}t||}t||}tj|jd|jd}d}}d}} } } |tjkrEt |}t |dd}t |\}}} t |\}} } ||||| || | |f S)z,Obtain statistics to calculate metric value.r)rONT)r0) r5rGrKr-tensorr2rOrrrVr^) r<r=rconcordant_pairsdiscordant_pairsn_total preds_ties target_ties preds_ties_p1 preds_ties_p2target_ties_p1target_ties_p2rrr_get_metric_metadataps*    rir`racon_min_dis_pairsrbrcrdc Cs|tjkr |||S|tjkrC||dd} |dur&tjd| j| jd}|dur4tjd| j| jd}| || |} |t| Stjdd|jD|j|jd} tjdd|jD|j|jd} t | | } d|| d| |dS) z-Calculate Kendall's tau from metric metadata.r6rWNrMcSg|]}t|qSrlenunique)rAprrrrCrDz"_calculate_tau..cSrlrrm)rAtrrrrCrD) rrrr-r_rNrOsqrtr/minimum)r<r=r`rarjrbrcrdrtotal_combinations denominator preds_unique target_unique min_classesrrr_calculate_taus     ryt_valuecCsl|}tjjtdg|tdg|}|}|}||}| |tjt d|j |j dS)zObtain p-value for a given Tensor of t-values. Handle ``nan`` which cannot be passed into torch distributions. When t-value is ``nan``, a resulted p-value should be alson ``nan``. rkrXnanrM) r- distributionsnormalNormalr_toisnan nan_to_numcdfwherefloatrNrO)rzrO normal_distis_nanp_valuerrr"_get_p_value_for_t_value_from_dists , "rrerfrgrhr'c Cs&||dd|d} |tjkrd|t| d} nQ||d} | |dur*|nd|dur2|ndd} | d|dur?|nd|durG|nd| 7} | |durT|nd|dur[|ndd| |d7} |t| } | tjkrxt| } | tjtjfvr| d 9} t| }| tjkr|d9}|S) z9Calculate p-value for Kendall's tau from metric metadata.r6rWrYNr rP) rrr-rrr#r(absr*r)rjrbrcrerfrdrgrhrr't_value_denominator_baserzmt_value_denominatorrrrr_calculate_p_values>  "   rr6 concat_preds concat_target num_outputscCs^|pg}|pg}t||t||||dkr!|d}|d}||||||fS)aUpdate variables required to compute Kendall rank correlation coefficient. Args: preds: Sequence of data target: Sequence of data concat_preds: List of batches of preds sequence to be concatenated concat_target: List of batches of target sequence to be concatenated num_outputs: Number of outputs in multioutput setting Raises: RuntimeError: If ``preds`` and ``target`` do not have the same shape r6)rrr9append)r<r=rrrrrr_kendall_corrcoef_updates      rc Cst|||\ }}}}}} } } } ||} t||||| | || | }|r.t| | |||| | | || nd}|jddkrE|}|durC|nd}|dd|fS)a/Compute Kendall rank correlation coefficient, and optionally p-value of corresponding statistical test. Args: Args: preds: Sequence of data target: Sequence of data variant: Indication of which variant of Kendall's tau to be used alternative: Alternative hypothesis for for t-test. Possible values: - 'two-sided': the rank correlation is nonzero - 'less': the rank correlation is negative (less than zero) - 'greater': the rank correlation is positive (greater than zero) Nrr6rP)riryrr2squeezeclamp)r<r=rr'r`rarcrerfrdrgrhrbrjtaurrrr_kendall_corrcoef_computesD  rrr$)rrrt_test)r$r%r&c Cst|tstdt|d|r|durtdtt|}|r)tt|nd}t||gg|j dkr7dn|j dd\}}t t |t |||\} } | durU| | fS| S)a Compute `Kendall Rank Correlation Coefficient`_. .. math:: tau_a = \frac{C - D}{C + D} where :math:`C` represents concordant pairs, :math:`D` stands for discordant pairs. .. math:: tau_b = \frac{C - D}{\sqrt{(C + D + T_{preds}) * (C + D + T_{target})}} where :math:`C` represents concordant pairs, :math:`D` stands for discordant pairs and :math:`T` represents a total number of ties. .. math:: tau_c = 2 * \frac{C - D}{n^2 * \frac{m - 1}{m}} where :math:`C` represents concordant pairs, :math:`D` stands for discordant pairs, :math:`n` is a total number of observations and :math:`m` is a ``min`` of unique values in ``preds`` and ``target`` sequence. Definitions according to Definition according to `The Treatment of Ties in Ranking Problems`_. Args: preds: Sequence of data of either shape ``(N,)`` or ``(N,d)`` target: Sequence of data of either shape ``(N,)`` or ``(N,d)`` variant: Indication of which variant of Kendall's tau to be used t_test: Indication whether to run t-test alternative: Alternative hypothesis for t-test. Possible values: - 'two-sided': the rank correlation is nonzero - 'less': the rank correlation is negative (less than zero) - 'greater': the rank correlation is positive (greater than zero) Return: Correlation tau statistic (Optional) p-value of corresponding statistical test (asymptotic) Raises: ValueError: If ``t_test`` is not of a type bool ValueError: If ``t_test=True`` and ``alternative=None`` Example (single output regression): >>> from torchmetrics.functional.regression import kendall_rank_corrcoef >>> preds = torch.tensor([2.5, 0.0, 2, 8]) >>> target = torch.tensor([3, -0.5, 2, 1]) >>> kendall_rank_corrcoef(preds, target) tensor(0.3333) Example (multi output regression): >>> from torchmetrics.functional.regression import kendall_rank_corrcoef >>> preds = torch.tensor([[2.5, 0.0], [2, 8]]) >>> target = torch.tensor([[3, -0.5], [2, 1]]) >>> kendall_rank_corrcoef(preds, target) tensor([1., 1.]) Example (single output regression with t-test) >>> from torchmetrics.functional.regression import kendall_rank_corrcoef >>> preds = torch.tensor([2.5, 0.0, 2, 8]) >>> target = torch.tensor([3, -0.5, 2, 1]) >>> kendall_rank_corrcoef(preds, target, t_test=True, alternative='two-sided') (tensor(0.3333), tensor(0.4969)) Example (multi output regression with t-test): >>> from torchmetrics.functional.regression import kendall_rank_corrcoef >>> preds = torch.tensor([[2.5, 0.0], [2, 8]]) >>> target = torch.tensor([[3, -0.5], [2, 1]]) >>> kendall_rank_corrcoef(preds, target, t_test=True, alternative='two-sided') (tensor([1., 1.]), tensor([nan, nan])) z>Argument `t_test` is expected to be of a type `bool`, but got .NzCArgument `alternative` is required if `t_test=True` but got `None`.r6rP)r) isinstancebool ValueErrortyperfrom_strr!r#rndimr2rr ) r<r=rrr'_variant _alternative_preds_targetrrrrrkendall_rank_corrcoefOs$ K  r)F)NNr6)N)rFr$)%typingrrrr-rtyping_extensionsr(torchmetrics.functional.regression.utilsrtorchmetrics.utilities.checksrtorchmetrics.utilities.datar r r torchmetrics.utilities.enumsr rr#tupler5rTr;rGrJrKrrVr^riryrrrrrrrrrs            (     .   ( ?