REVIEW 7 cited by
Jacobian Descent for Multi-Objective Optimization
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Jacobian Descent for Multi-Objective Optimization
read the original abstract
Many optimization problems require balancing multiple conflicting objectives. As gradient descent is limited to single-objective optimization, we introduce its direct generalization: Jacobian descent (JD). This algorithm iteratively updates parameters using the Jacobian matrix of a vector-valued objective function, in which each row is the gradient of an individual objective. While several methods to combine gradients already exist in the literature, they are generally hindered when the objectives conflict. In contrast, we propose projecting gradients to fully resolve conflict while ensuring that they preserve an influence proportional to their norm. We prove significantly stronger convergence guarantees with this approach, supported by our empirical results. Our method also enables instance-wise risk minimization (IWRM), a novel learning paradigm in which the loss of each training example is considered a separate objective. Applied to simple image classification tasks, IWRM exhibits promising results compared to the direct minimization of the average loss. Additionally, we outline an efficient implementation of JD using the Gramian of the Jacobian matrix to reduce time and memory requirements.
Forward citations
Cited by 7 Pith papers
-
Learn from your own latents and not from tokens: A sample-complexity theory
Latent prediction SSL recovers latent trees from PCFG data with sample complexity constant in hierarchy depth L (up to logs), unlike exponential for token-level or supervised methods.
-
Multi-objective design of photon blockade for bright single-photon sources
A new multi-objective optimization framework using Liouville-space adjoint formulation and simulated annealing achieves nearly 60% success rate in designing photon blockade with g2(0) < 0.1 and bounded brightness.
-
Multi-Objective Reference-Aligned Machine Unlearning
RAUL is a multi-objective unlearning framework using bounded KL alignment to a reference distribution and Jacobian descent that reports closer performance to full retraining than single-objective baselines.
-
A Unified Framework for Gradient Aggregation in Multi-Objective Optimization
A unifying framework for gradient aggregation in multi-objective optimization establishes convergence to Pareto stationarity under a sufficient alignment condition and introduces capped MGDA for robustness in adversar...
-
When Losses Align: Gradient-Based Composite Loss Weighting for Efficient Pretraining
A bilevel method learns composite pretraining loss weights online via gradient alignment with a downstream objective, matching tuned baselines at roughly 30% extra cost over one training run.
-
Delve into the Applicability of Advanced Optimizers for Multi-Task Learning
APT augments multi-task learning by adapting advanced optimizers via momentum balancing and light direction preservation, delivering performance gains on four standard MTL datasets.
-
Accessing the performance of CC2 for excited state dynamics: a benchmark study with pyrazine
RI-CC2 simulations of pyrazine internal conversion match the experimental 22 fs decay time, identify Q9a and Q8a modes as drivers, and show the dark A1u state participates actively.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.