Metrics

weavenet.metric.binarize(m)

Binarizes each matrix in a batch into the one-to-one format (if N=M). If N>M, N-M vertices will have no partner and vice versa.

Shape:

• m: \((B, N, M)\)

• output: \((B, N, M)\)

Parameters

m (Tensor) – a continously-relaxed assignment between sides \(a\) and \(b\), where |a|=N, |b|=M.

Returns

A binarized batched matrices.

weavenet.metric.is_one2one(m)

Checks whether each matrix in a batch m has no duplicated correspondence.

Shape:

• m: \((B, N, M)\)

• output: \((B)\)

Parameters

m (Tensor) – a binary assignment between sides \(a\) and \(b\), where |a|=N, |b|=M.

Returns

A binary bool vector.

weavenet.metric.is_stable(m, sab, sba_t)

Checks whether each matrix in a batch m is a stable match or not.

Shape:

• m: \((B, N, M)\)

• sab: \((B, N, M)\)

• sab: \((B, M, N)\)

• output: \((B)\)

Parameters

• m (Tensor) – a binary (or continously-relaxed) assignment between sides \(a\) and \(b\), where |a|=N, |b|=M.

• sab (Tensor) – a satisfaction at matching of agents in side \(a\) to side \(b\).

• sba – a satisfaction at matching of agents in side \(b\) to side \(a\).

• sba_t (Tensor) –

Returns

A binary bool vector.

Return type

Tensor

weavenet.metric.count_blocking_pairs(m, sab, sba_t)

Counts the number of blocking pairs for each matrix in batch m.

Shape:

• m: \((B, N, M)\)

• sab: \((B, N, M)\)

• sab: \((B, M, N)\)

• output: \((B)\)

Parameters

• m (Tensor) – a binary (or continously-relaxed) assignment between sides \(a\) and \(b\), where |a|=N, |b|=M.

• sab (Tensor) – a satisfaction at matching of agents in side \(a\) to side \(b\).

• sba – a satisfaction at matching of agents in side \(b\) to side \(a\).

• sba_t (Tensor) –

Returns

A count vector.

Return type

Tensor

weavenet.metric.sexequality_cost(m, cab, cba_t, pformat=PreferenceFormat.cost)

Calculates sexequality costs.

Shape:

• m: \((B, N, M)\)

• cab: \((B, N, M)\)

• cab: \((B, M, N)\)

• output: \((B)\)

Parameters

• m (Tensor) – a binary (or continously-relaxed) assignment between sides \(a\) and \(b\), where |a|=N, |b|=M.

• cab (Tensor) – a cost at matching of agents in side \(a\) to side \(b\).

• cba – a cost at matching of agents in side \(b\) to side \(a\).

• cba_t (Tensor) –

• pformat (PreferenceFormat) –

Returns

A cost vector.

Return type

Tensor

weavenet.metric.egalitarian_score(m, cab, cba_t, pformat=PreferenceFormat.cost)

Calculates egalitarian score.

Shape:

• m: \((B, N, M)\)

• cab: \((B, N, M)\)

• cab: \((B, M, N)\)

• output: \((B)\)

Parameters

• m (Tensor) – a binary (or continously-relaxed) assignment between sides \(a\) and \(b\), where |a|=N, |b|=M.

• cab (Tensor) – a cost at matching of agents in side \(a\) to side \(b\).

• cba – a cost at matching of agents in side \(b\) to side \(a\).

• cba_t (Tensor) –

• pformat (PreferenceFormat) –

Returns

A score vector.

Return type

Tensor

weavenet.metric.balance_score(m, cab, cba_t, pformat=PreferenceFormat.cost)

Calculates egalitarian score.

Shape:

• m: \((B, N, M)\)

• cab: \((B, N, M)\)

• cab: \((B, M, N)\)

• output: \((B)\)

Parameters

• m (Tensor) – a binary (or continously-relaxed) assignment between sides \(a\) and \(b\), where |a|=N, |b|=M.

• cab (Tensor) – a cost at matching of agents in side \(a\) to side \(b\).

• cba – a cost at matching of agents in side \(b\) to side \(a\).

• cba_t (Tensor) –

• pformat (PreferenceFormat) –

Returns

A score vector.

Return type

Tensor

weavenet.metric.calc_all_fairness_metrics(m, cab, cba_t, pformat=PreferenceFormat.cost)

Calculates the three fairness scores (sex-equality, egalitarian score, and balance score).

Shape:

• m: \((B, N, M)\)

• cab: \((B, N, M)\)

• cab: \((B, M, N)\)

• output: \((B)\)

Parameters

• m (Tensor) – a binary (or continously-relaxed) assignment between sides \(a\) and \(b\), where |a|=N, |b|=M.

• cab (Tensor) – a cost at matching of agents in side \(a\) to side \(b\).

• cba – a cost at matching of agents in side \(b\) to side \(a\).

• cba_t (Tensor) –

• pformat (PreferenceFormat) –

Returns

The three score vectors.

Return type

Tuple[Tensor, Tensor, Tensor]