{"paper_id":"2601.07942","title":"Enhancing Portfolio Optimization with Deep Learning Insights","authors":["Brandon Luo","Jim Skufca"],"paper_date":"2026-01-12","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2601.07942v1","page_url":"https://thequant.space/flowcharts/enhancing-portfolio-optimization-with-deep-learning-insights/","scores":{"empirical_rigor":6,"math_complexity":7,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes training deep‑learning portfolio models with limited regime data using pre‑training techniques and transformer architectures that incorporate state variables.","quote":"We propose training models with limited regime data using pre‑training techniques and leveraging transformer architectures for state variable inclusion.","summary":"Our work focuses on deep learning (DL) portfolio optimization, tackling challenges in long-only, multi-asset strategies across market cycles. We propose training models with limited regime data using pre-training techniques and leveraging transformer architectures for state variable inclusion.…","strategy":"Transformer portfolio model"},"assumptions":{"status":"available","items":[{"assumption":"Daily trading adjustments are possible with negligible slippage.","type":"execution","stated":"implied","quote":"Considering the active trading nature of our LSTM model, like the approach used in the Oxford study, we accounted for a 0.01% transaction fee on daily adjustments in allocation weights to determine net returns."},{"assumption":"Missing daily data points can be filled by forward‑filling and back‑filling.","type":"data","stated":"explicit","quote":"To ensure consistency in daily (trading days) data points, we employed forward filling and backfilling techniques."},{"assumption":"Economic indicators reported in units can be transformed to year‑over‑year percent changes for comparability.","type":"data","stated":"explicit","quote":"we transformed twelve economic indicators reported in units into year‑over‑year percent changes, ensuring a standardized comparison across the dataset."},{"assumption":"Pre‑training on a broader historical dataset improves transfer learning for under‑represented regimes.","type":"model","stated":"explicit","quote":"The Oxford study's data limitations led us to hypothesize that pretraining the model on broader historical data from 1986 could improve transfer learning, especially for less represented scenarios in the 2006‑2023 period."},{"assumption":"L2 regularization and dropout mitigate overfitting in the transformer model.","type":"model","stated":"explicit","quote":"To enhance model robustness, we added L2 Regularization to the Adam optimizer and included a dropout layer for normalization and to counter overfitting."}]},"datasets":{"status":"available","items":[{"name":"Oxford dataset","provider":null,"kind":"macro","access":"unknown","period":"2006‑2023","universe":"multi‑asset long‑only portfolio (VTI, AGG, DBC, ^VIX, etc.)","quote":"The Oxford study's data limitations led us to hypothesize that pretraining the model on broader historical data from 1986 could improve transfer learning, especially for less represented scenarios in the 2006‑2023 period."},{"name":"Pre‑training dataset","provider":null,"kind":"market-prices","access":"unknown","period":"1986‑present","universe":"stock and bond price data (Wilshire and BofA), VTI, AGG, ^VIX, DBC","quote":"Our pre‑training dataset included stock and bond price data from our asset experiments (Wilshire and BofA) dating back to 1986, for initializing weights in VTI and AGG funds."},{"name":"Economic indicators (21 new features)","provider":"FRED / Yahoo / Kaggle","kind":"macro","access":"public","period":null,"universe":null,"quote":"These features comprised two financial market indicators and 19 economic indicators (Figure 1).","known_id":"fred"}],"known":[{"id":"cboe","name":"Cboe","provider":"Cboe Global Markets","kind":"options","access":"commercial","quote":"Notably, we substituted the absence of an ETF for VIX with the daily closing values of the CBOE Volatility Index (^VIX) in our dataset.","mentions":2,"used_in":1},{"id":"kaggle","name":"Kaggle","provider":"Kaggle","kind":"other","access":"public","quote":"Three of these instruments' data were sourced from the Federal Reserve Banks’ website (fred.stlouisfed.org), while gold prices were obtained from Kaggle (kaggle.com) Figure 1 - Data table Figure 2 - Correlations 3.2.3 Features experiment Our third experiment sought to evaluate whether incorporating additional features beyond security price and return, such as financial market and economic indicators,","mentions":2,"used_in":1},{"id":"yahoo-finance","name":"Yahoo Finance","provider":"Yahoo","kind":"market-prices","access":"public","quote":"It is important to note that the data was unnormalized and downloaded directly from Yahoo Finance's API.","mentions":1,"used_in":1}],"sample":{"start":null,"end":null,"market":"US equities, bonds, commodities and volatility index","asset_class":"multi-asset","frequency":"daily"},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"unclear","headline_net_of_costs":false,"detail":null,"quote":null,"source":"conflict","rule":"R4b","candidates":{"front_matter":"partial","evidence_card":"yes","evidence_card_quoted":true},"conflict_kind":"graded","card_detail":"0.01% transaction fee on daily adjustments to allocation weights","card_quote":"Considering the active trading nature of our LSTM model, like the approach used in the Oxford study, we accounted for a 0.01% transaction fee on daily adjustments in allocation weights to determine net returns.","meaning":"the platform classifier and the paper-text extraction disagree, so no value is asserted"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"The replicated LSTM model achieves a Sharpe ratio of 1.828 (net of transaction costs), closely matching the Oxford study's 1.858 and outperforming MVO and Balanced benchmarks in annual return, profitability, and max drawdown.","kind":"performance","quote":"Our replicated portfolio's Sharpe ratio closely aligns with the Oxford study (1.828 vs 1.858), demonstrating superior returns in comparison to the MVO and Balanced portfolios in terms of annual return, profitability (percent of positive returns), and risk metrics (max drawdown)."},{"finding":"Training and validation curves show most validation Sharpe peaks between 20 and 40 epochs, with some models improving up to epoch 100, suggesting early stopping to avoid overfitting.","kind":"statistical","quote":"These curves revealed five distinct training patterns, with most validation curves peaking in Sharpe ratios between 20 and 40 epochs. One curve improved until epoch 100, while another peaked around epoch 90, suggesting the Oxford team might have stopped training at epoch 100 for generalization and to avoid overfitting."},{"finding":"Dynamic asset allocations vary widely: VTI weights from 0.1‑0.6, AGG from 0.2‑0.9, DBC from 0.0‑0.07, and ^VIX from 0.0‑0.25 across time.","kind":"performance","quote":"Figure 6 shows the dynamic nature of our LSTM model in reallocating capital across asset classes, with significant variations in allocations to VTI (0.1 - 0.6) and AGG (0.2 - 0.9), while allocations to DBC and ^VIX ranged from 0.0 to 0.07 and 0.0 to 0.25, respectively."}],"reported_performance":{"strategy":"Transformer portfolio model","sharpe":2.7,"sharpe_benchmark":0.8,"annual_return_pct":null,"total_return_pct":null,"max_drawdown_pct":null},"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":30,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T17:16:40Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}