Top-K Multi-Positive Contrastive Objective

• Top-K Multi-Positive Contrastive Objective (MSCL) is a contrastive learning framework that leverages multiple positive samples to improve recommendation quality.
• It modifies the classic InfoNCE loss with importance-aware weighting, balancing the influence of positive and negative interactions.
• Empirical results on datasets like Yelp2018 and Amazon-Book show that MSCL achieves higher accuracy and faster convergence compared to traditional methods.

The Top-K Multi-Positive Contrastive Objective (MSCL) is a contrastive learning framework tailored for recommender systems under Top-K recommendation metrics. In this paradigm, MSCL modifies the classic InfoNCE/NT-Xent loss, incorporating a strategy that samples multiple positive items per user and applies a tunable importance weighting between positive and negative terms. This combination enhances both the utilization of sparse user-item interactions and the balance of gradient signal, resulting in improved recommendation accuracy and efficient optimization, particularly on sparse datasets (Tang et al., 2021).

1. Formalization of the Top-K Recommendation Task

The Top-K recommendation setting operates on the bipartite interaction graph $G = (U \cup I, E^+)$, where $U$ denotes the set of users, $I$ the set of items, and $E^+ \subseteq U \times I$ the observed positive user-item interactions. For each user $u \in U$, the system must learn an embedding $e_u \in \mathbb{R}^d$, and similarly for each item $i \in I$ an embedding $e_i \in \mathbb{R}^d$. Recommendation is generated by ranking items for $u$ according to the predicted score $\hat{y}_{u i} = e_u^T e_i$, selecting the K items with the highest scores.

This formalism is central to embedding-based collaborative filtering and underpins both the classical contrastive baseline and the MSCL framework (Tang et al., 2021).

2. Standard Contrastive Loss (CL) and Its Limitations

The NT-Xent loss—prevalent in models such as SimCLR and GraphCL—adapts to recommendation by contrasting a user’s positive item against negatives drawn from the remainder of the minibatch. Given a minibatch $U$0 of size $U$1, the positive pair $U$2 is contrasted with $U$3. The loss is:

$U$4

where $U$5 is the cosine similarity of $U$6 and $U$7, and $U$8 is the temperature.

The batch loss is averaged over $U$9 such pairs. A critical limitation is the severe imbalance for each user: a single positive term against $I$0 negatives, which is further exacerbated by the sparsity common in practical recommendation scenarios. Each training step exploits only one positive item, leading to under-utilization of sparse user-item interaction signals (Tang et al., 2021).

3. Derivation and Formulation of the Multi-Positive Contrastive Loss (MSCL)

3.1 Importance-aware Contrastive Loss (ICL)

ICL introduces a weighting parameter $I$1 to control the relative contribution of positive and negative terms:

$I$2

Setting $I$3 recovers the symmetric NT-Xent loss, while $I$4 increases the emphasis on positives—helpful in extremely sparse data.

3.2 Multi-Positive Contrastive Loss (MCL) via Data Augmentation

Instead of a single positive sampled per user, MCL samples $I$5 distinct positives $I$6 from the user’s interaction history (without replacement). The per-user loss is aggregated over all sampled positives:

$I$7

3.3 Final Multi-Sample Contrastive Loss (MSCL)

Combining importance weighting and multi-positive sampling, MSCL is formally written as:

$I$8

with the overall minibatch loss:

$I$9

This approach enables simultaneous utilization of multiple positives, and the weighting scheme directly modulates the gradient flow to address imbalance (Tang et al., 2021).

4. Multi-Positive Sampling and Data Augmentation

At each iteration, for each user $E^+ \subseteq U \times I$0 in the minibatch, let $E^+ \subseteq U \times I$1 denote all known positives. MSCL samples $E^+ \subseteq U \times I$2 positives without replacement, resulting in $E^+ \subseteq U \times I$3 potential positive sets and thus significantly augments the sampling space. Each sampled $E^+ \subseteq U \times I$4 receives a parallel contrastive loss with the same set of negatives, substantially increasing supervisory signal—even in the case of very short user histories. This combinatorial data augmentation effect is a defining characteristic of MSCL, enabling the model to effectively utilize limited positive interactions (Tang et al., 2021).

5. Hyperparameterization: Balancing Positives and Negatives

The hyperparameter $E^+ \subseteq U \times I$5 is the central knob for controlling the positive/negative trade-off. $E^+ \subseteq U \times I$6 yields the conventional contrastive loss, but tuning is dataset-dependent: $E^+ \subseteq U \times I$7 works best for denser sets, and $E^+ \subseteq U \times I$8 for ultra-sparse domains (e.g., Alibaba-iFashion). Selection of $E^+ \subseteq U \times I$9, the number of positive views, is typically in the range $u \in U$0, further increasing as user positive history and dataset size permit (Tang et al., 2021).

6. Integration with Graph Encoder Architectures

MSCL is agnostic to the embedding-based encoder and is demonstrated with LightGCN and sLightGCN. User and item embeddings $u \in U$1 and $u \in U$2 are computed as the layer-wise averages over GCN propagation steps:

$u \in U$3

The cosine similarity between $u \in U$4 and $u \in U$5, $u \in U$6, is input to the MSCL objective, replacing prior ranking-based losses such as BPR (Tang et al., 2021).

7. Empirical Findings and Advantages in Sparse Regimes

Benchmarking on Yelp2018, Amazon-Book, and Alibaba-iFashion datasets (interaction densities $u \in U$7, $u \in U$8, $u \in U$9, respectively), MSCL with $e_u \in \mathbb{R}^d$0 and tuned $e_u \in \mathbb{R}^d$1 significantly outperforms the single-positive CL baseline:

Dataset	Metric	CL	MSCL	Relative Improvement
Yelp2018	recall@20	0.0655	0.0691	+5.0% (NDCG@20)
Amazon-Book	recall@20	0.0480	0.0580	+17% (NDCG@20)
Alibaba-iFashion	recall@20	0.1152	0.1201	+4% (NDCG@20)

These gains are accompanied by marked improvements in convergence speed ($e_u \in \mathbb{R}^d$250 epochs for MSCL versus $e_u \in \mathbb{R}^d$3900 for BPR) and nearly identical per-epoch computational cost on modern GPUs (Tang et al., 2021).

In summary, MSCL operationalizes combinatorially-augmented contrastive learning, tuned by explicit positive:negative weighting. Its principal benefit lies in overcoming the dual challenges of imbalance and sparse data endemic to practical Top-K recommender systems (Tang et al., 2021).