context stringlengths 102 1.98k | A stringlengths 104 1.61k | B stringlengths 100 1.73k | C stringlengths 105 1.82k | D stringlengths 102 1.47k | label stringclasses 4
values |
|---|---|---|---|---|---|
However, in spite of this well-documented disparity, most existing modeling approaches, including those designed to strengthen smart grid integration, have not meaningfully incorporated social equity concerns. A recent review of data driven grid resilience strategies highlights this oversight, pointing out that many pr... | On the other hand, a promising modeling framework, named Markov Decision Process (MDP) shows promising potential in optimizing energy grids. MDP has been used in many fields, including robotics, supply chain management, and healthcare, to make sequential decisions in environments with probabilistic transitions and rewa... | MDPs have become increasingly useful to modeling decision-making under uncertainty within power systems, especially with the incorporation of renewables. Their capacity to handle stochastic dynamics over time makes them well-suited for applications where energy demand, generation, and storage are subject to variability... | International energy data reveal that nearly 3,000 gigawatts of renewable energy projects are awaiting grid connection, underscoring how grid capacity has become a bottleneck in the deployment of renewables [2]. While investments in renewable energy generation have nearly doubled since 2010, funding directed toward gri... | The integration of renewable energy sources such as wind and solar into existing power grids is currently hindered by significant infrastructure limitations. Traditional power grids were primarily designed to accommodate dispatchable energy sources like fossil fuels, which provide consistent and controllable output. In... | B |
Before proceeding to Proposition 6 we need to formalize the use of non-contractual cashflows in calculating policy values and other quantities. | under Scandinavian-style regulation, the valuation is both net premium and gross premium, since τtL=πtL=Pt\tau_{t}^{L}=\pi_{t}^{L}=P_{t}. | Valuation premium rates τtL≠Pt\tau_{t}^{L}\not=P_{t} introduce non-contractual cashflows. Non-contractual cashflows also arise often other calculations. | Bonus: Any distribution of surplus to with-profit policies involves non-contractual cashflows. Bonus is considered in Section 13. | Before proceeding to Proposition 6 we need to formalize the use of non-contractual cashflows in calculating policy values and other quantities. | B |
The remainder of this paper is structured as follows: Section II introduces the wind-electrolyser system under study, Section III details the techno-economic framework, Section IV elaborates on the green hydrogen use cases, Section V shows the results and the LCOH sensitivity analysis, while Section VI concludes this w... | The climate emergency and ambitious net-zero targets position hydrogen as a key alternative to replace gas and fossil fuels, especially in difficult to decarbonise sectors via electrification, such as aviation, freight transportation, and heavy industry [1]. Many countries have established roadmaps and policy to suppor... | where PH2,tP_{H2,t} is the power used for hydrogen production, PExp,tP_{Exp,t} the power exported to the grid and PCurt,tP_{Curt,t} the power curtailed, due to restricted grid access or network constraints. PH2,tP_{H2,t} is used to supply both the electrolyser and the compressor/storage system and may be supplie... | However, there is a lack of understanding on how to evaluate its potential due to the variety of deployment options and business cases, which include different physical configurations, commercial arrangements, location, operation strategies, and interplay with electricity markets and regulatory frameworks. To address ... | The analysis in this paper is based on a reference system comprising a 10 MW wind plant and a proton exchange membrane electrolyser (PEMEL) illustrated in Fig. 1. PEMEL was selected because of its advantages compared to other technologies (alkaline and solid oxide), including high current density, fast response, and hi... | D |
Takakura, Jun’ya, Shinichiro Fujimori, Naota Hanasaki, Tomoko Hasegawa, Yukiko Hirabayashi, Yasushi Honda, Toshichika Iizumi, et al. 2019. | Tokarska, Katarzyna B., Martin B. Stolpe, Sebastian Sippel, Erich M. Fischer, Christopher J. Smith, Flavio Lehner, and Reto Knutti. 2020. | Tokarska, Katarzyna B., Martin B. Stolpe, Sebastian Sippel, Erich M. Fischer, Christopher J. Smith, Flavio Lehner, and Reto Knutti. 2020. | Tokarska, Katarzyna B., Martin B. Stolpe, Sebastian Sippel, Erich M. Fischer, Christopher J. Smith, Flavio Lehner, and Reto Knutti. 2020. | Tokarska, Katarzyna B., Martin B. Stolpe, Sebastian Sippel, Erich M. Fischer, Christopher J. Smith, Flavio Lehner, and Reto Knutti. 2020. | A |
For illustration, Fig. 1 - 6 present the weekly cumulative out-of-sample returns for portfolios employing the (NoM), (RoM-RKH) models, alongside BMP and EQP, across each dataset. The out-of-sample returns from each rolling window are consolidated into a single out-of-sample return series, creating these plots. The cumu... | From Fig. 1, we note that RoMP with AA = 6 gives maximum rewards compared to NoMP in the Nikkei market. And NoMP with AA = 6 is more rewarding than BMP and EQP. | The EQP outperformed other portfolios from Fig. 6. However, the RoMP with AA = 15 gives better cumulative returns than NoMP and BMP. | From Fig. 3, in the NIFTY 50 market, RoMP with AA = 3 gives maximum rewards; NoMP with AA = 9 is suboptimal to EQP. However, it gives returns nearly equal to the BMP. | From Fig. 2, the RoMP with AA = 9 gives the maximum reward in the S&P market. However, the NoMP with AA = 9 is suboptimal compared to the BMP and EQP. EQP gives a nearly equal return as RoMP with AA = 9. | A |
To be clear, we call zjtz_{jt} the measured productivity, ωjt\omega_{jt} the total factor productivity in the context of our model, and y~jt=θ^lnyjt\tilde{y}_{jt}=\widehat{\theta}\ln y_{jt} the contribution of top-worker quality to measured productivity. The estimate of output elasticity θ^\widehat{\theta} is sec... | Taking into account the optimal matching between top workers and others, we estimate the production function for Canadian firms over the period 2003-2015, from which we obtain the firm-level total factor productivity (Hicks-neutral technology). We then obtain the measured total factor productivity which consists of the... | Second, the negative contribution of the top-worker quality to the slowdown of measured productivity growth is associated with the falling aggregate top-worker quality since the output elasticity θ^\widehat{\theta} does not vary over time. The quality of top workers dropped more than that of non-top workers. On average... | The quality of non-top workers also declined from 2003 to 2015, but to a lesser extent when contrasted with top workers. This resulted in a narrowing quality gap between top workers and the average non-top workers, we call the gap the match efficiency. Thus, the declined in match efficiency, considered exogenous in our... | The measured aggregate productivity is slightly different between the two indexes, and we use the measure based on the top-worker quality for productivity analysis. Thus, we must make it clear that although the subsequent analysis discusses the role of top workers in measured productivity growth, it is equivalent to th... | D |
The original system of differential equations contains seven numerical constants. Their traditional identification is problematic due to the typical severe information shortages; see e.g. [50]. However, when using trend quantifiers, numerical multiplicative constants such as A>0A>0 or B<0B<0 can generally be eliminated... | The following set of relations [44] illustrates the flexibility of trend-based analysis grounded in (1): | The autonomous system of first-order differential equations given in (8), adapted from [7], is used as a model of rumour spreading. | The system of numerically based differential equations (8) is translated into the trend-based RRM (11) using the conversion rules given in (9) and (10). | Moreover, the following simple expressions are used to apply trend-based analysis to the system of differential equations; for details, see e.g. [25, 35]: | D |
To investigate these aspects, we turn to the Generalized Hurst Exponent H(q)H(q), which extends the classical Hurst analysis to capture multiscaling behavior and fluctuation heterogeneity. Specifically, by analyzing H(q)H(q) for different values of the moment order qq, we can assess whether the P/E ratio dynamics fol... | To probe the scaling properties and memory characteristics of the P/E ratio, we compute the Generalized Hurst Exponent H(q)H(q) for q=1q=1 to 55. A decreasing H(q)H(q) with increasing qq signals multifractality—heterogeneous scaling behavior across fluctuation magnitudes. This approach allows us to quantify how the p... | To investigate these aspects, we turn to the Generalized Hurst Exponent H(q)H(q), which extends the classical Hurst analysis to capture multiscaling behavior and fluctuation heterogeneity. Specifically, by analyzing H(q)H(q) for different values of the moment order qq, we can assess whether the P/E ratio dynamics fol... | Our analysis began with the recognition that elevated P/E ratios alone are insufficient to reliably predict drawdowns in Nifty 50 price movements. This motivated a deeper investigation into the probabilistic structure of returns across multiple time horizons, leading to the study of return distributions through empiric... | To refine this picture, we consider the Tsallis entropy, which allows sensitivity to the statistical weight of rare versus frequent events through the entropic index qq. For q=0.1q=0.1, the entropy value of 0.92 emphasizes the influence of rare events, while for q=2q=2, the higher value of 0.98 highlights the dominance... | A |
The data used for both training and evaluation is generated using the synthetic environment first introduced in wilinski2025classifying . | In order to find the optimal parameters, the model is trained by minimizing cross-entropy between the mentioned distribution and the true labels. | Four groups of chartists are further divided according to their strategy (trend following, and mean reversion). | It is worth noting that the resulting distribution is very symmetric, even for values which are far from the ground truth. | The setting is the same as in the mentioned article, which means there are 1590 agents divided into 15 groups. | D |
In our case, the American option terminates early if a drawdown event occurs. Specifically, this means the option is exercised either at a random stopping time or at the first time when the drawdown of the stock price exceeds a fixed threshold, whichever comes first. | Random termination is a common feature of many financial products. For example, the random termination moment could correspond to the default time of a company (see, e.g., [2, p. 27] or [32]) or an asset-price-independent time cap following an exponential or Erlang distribution (see [4]). The latter case, which leads t... | Explicitly including protection against significant drawdowns in financial contracts is a common practice aimed at minimizing potential losses for the seller. Therefore, the list of papers addressing contracts that incorporate drawdown or drawup feature is quite long; see, | Notably, when x¯\overline{x} exceeds log(K)+c\log(K)+c, the option becomes worthless unless x<log(K)x<\log(K). This is because, at such a level of x¯\overline{x}, the stock price cannot reach the strike price before the option is terminated by the drawdown trigger. Of course, one can hypothetically consider a pair (x... | In our case, the American option terminates early if a drawdown event occurs. Specifically, this means the option is exercised either at a random stopping time or at the first time when the drawdown of the stock price exceeds a fixed threshold, whichever comes first. | B |
Calibration results for the one–week horizon are summarised in Table 6. The pp–values are exceedingly small for most models, again reflecting the large sample size and heavy tails in equity returns. The Merton model exhibits a comparatively large pμp_{\mu}, consistent with its inability to adapt to time–varying risk. N... | Figures 3 and 4 visualise the relative improvements for the one–week horizon. NeuralLevy exhibits strong gains in both metrics, particularly in CRPS, underscoring the advantage of capturing state–dependent jumps at longer horizons. | Results for the five–day horizon are presented in Table 4. NeuralLevy remains the top performer, achieving a 3.60% improvement in log score and a substantial 14.36% improvement in CRPS relative to the best baseline. The diffusion–only NN serves as the strongest baseline at this horizon but still underperforms NeuralLev... | Figures 1 and 2 illustrate the relative improvements graphically. NeuralLevy provides positive gains in both log score and CRPS relative to the best baseline, whereas competing methods exhibit negative improvements in at least one metric. | Table 4: Core predictive metrics for the one–week (1W) horizon. Relative improvements are computed relative to the best baseline; positive values denote better performance. | A |
Of course, legislation is only part of the remedy. Just as financial firms tend to game stress tests (Board of Governors of the Federal Reserve | (ii) Scope. The introduction of AI agents into production might also lead firms to expand into new industries. There are at least two mechanisms through which this could happen. The first is technical: AI agents might become quite good at transfer learning—their training and expertise in one domain might generalize to ... | For these reasons, governments and the academic researchers that can contribute public domain knowledge to regulatory efforts will need access to information that is now considered proprietary to the firm. Regulators in other domains, of course, routinely gain access to confidential information: pharmaceutical firms ha... | System, 2016), AI firms might have considerable leeway to manipulate the information they share, or to flout safety procedures when it conflicts with profit motives. Thus, even if the legal boundaries of firms are made porous, this raises new economic questions about when to inspect and what to look for.181818Varas | Such identity and registration infrastructure are currently missing for AI agents (Hadfield, 2025; Chan et al., 2025). Building them out will be essential, but their design raises questions around legal accountability. One possible route is to require that any AI agent entering into a contract or transaction be registe... | C |
The spillover index has been used in many studies to analyze the linkages between financial markets (e.g., [23, 3, 7, 6, 1, 16, 18, 19, 5, 21]). However, although there is a large literature addressing spillover effects, relatively few studies explicitly consider the role of noise in financial markets or seek to separa... | In this respect, the distinction between noise and signal not only helps us to better understand the volatility spillovers, but also allows us to identify spillovers caused by noise-aggregated, erratic information that does not lead to consistent and reliable results. Recent advances in deep learning, particularly the ... | Furthermore, the process of denoising significantly diminishes noise and improves the alignment between spillover dynamics and actual events, such as financial crises or unique shocks. This enhancement in methodology not only bolsters the reliability of spillover analysis but also equips market participants and policym... | The spillover index has been used in many studies to analyze the linkages between financial markets (e.g., [23, 3, 7, 6, 1, 16, 18, 19, 5, 21]). However, although there is a large literature addressing spillover effects, relatively few studies explicitly consider the role of noise in financial markets or seek to separa... | Therefore, the first contribution of this study is to show the effect of denoising when examining the impact of spillovers across markets. By comparing results with and without neural network-based denoising, we provide direct evidence of the importance of noise removal in spillover analysis. Thus, return and volatilit... | A |
Technology Access and Infrastructure: Beyond rate design modifications, complementary technology and assistance programs are essential for equitable variable pricing implementation. Subsidized access to smart thermostats, programmable appliances, and battery storage can democratize demand flexibility capabilities, whil... | Consumer price sensitivity plays a crucial role in both conditions of Proposition 7. The ratio Al/Ah>1A_{l}/A_{h}>1 acts as a multiplier that amplifies the relative impact of pricing changes on low-income consumers’ utility. Because low-income consumers have a higher marginal disutility of expenditure (Al>AhA_{l}>A_{h}... | Our analysis uses utility rather than consumer surplus as the primary welfare measure to capture broader impacts of electricity access on household well-being. Consumer surplus, calculated by normalizing utility by the marginal disutility of expenditure (AiA_{i}), measures welfare changes in terms of willingness to pay... | Targeted Protection Measures: Demand response initiatives targeting vulnerable populations should prioritize measures not directly tied to demand response, such as insulation investments and assistance with healthcare during temperature extremes. This should be done not just because of fairness concerns, but because it... | These results suggest two sets of policies for different consumer types. Policies that directly promote demand responsiveness, such as smart meter deployment, time-varying rate education, or flexible appliance adoption, should be prioritized for consumers who lack flexibility due to technical limitations. However, a di... | C |
We conducted roll-forward experiments as follows. For each week, for each clustering method, we: 1) Reset (or recompute) the sector or cluster assignment for each stock as of the end of the week. 2) Compute the corresponding linear model weightings. 3) Compute errors against the linear models each day over the followin... | Out-of-Sample prediction and error measurement: In evaluation, out of sample, we use the estimated loadings to predict each stock’s return in the following period. We compute the prediction error each day as the difference between actual and predicted returns. We summarize performance using both root mean squared error... | Table 1: Average Daily Out-of-Sample Prediction Error by Clustering Method (Sorted by RMSE, January 2022 to December 2024. Lower numbers are better.) Note that price-based returns methods dominate. | We compare average out-of-sample prediction errors across clustering methods to assess which produces the most informative Summary of cross-sectional return variation. | Prediction Error We compare the predictive performance of each clustering method by evaluating the out-of-sample return prediction errors using the linear factor model. We report each method’s average root mean squared error (RMSE) and mean absolute error (MAE) across the whole evaluation period. Lower values indicate ... | D |
={BtPt∗<E[Pt+1],AtPt∗>E[Pt+1],Nothingelse\displaystyle=\begin{cases}B_{t}&P_{t}^{\ast}<E[P_{t+1}],\\ | OjO_{j} operates as a collector of bids and asks and a matchmaker that matches and settles the best bid and ask orders in a form of a trade. The best bid price is the highest bid price available in the market at time tt or Ptb=sup{Pb:(Pbi,qbi)∈Bt}P_{t}^{b}=\sup\{P_{b}:(P^{i}_{b},q^{i}_{b})\in B_{t}\}, while the best as... | This means that the trader submits an ask order if his private expectation of the price-change is a drop and a bid order if the price is expected to rise. The price expectations are created as | Once submitted orders can not be modified. The orderbook, as a match-maker, iteratively matches best bids with best asks, that is the highest-priced bid and the lowest-priced ask, until such pairs can be formed and settles them at the mid-price of the bid’s and the ask’s price. Any unmatched order at the end of the tim... | A trade is settled at the mid-price P(A,B)=(PA+PB)/2P_{(A,B)}=(P_{A}+P_{B})/2. For the trader ii this means that the change in cash ΔMi,t\Delta M_{i,t} and change in stock holdings ΔSj,i,t\Delta S_{j,i,t} are promptly modified as | B |
The hyperparameter β\beta controls the relative weighting of the latent variable 𝐳\mathbf{z} term in the objective function. In one-, three-, and four-feature control experiments, we observe that smaller values of β\beta enable the model to more effectively capture residual shape features. when β\beta is too large, th... | We begin by controlling three features: the level (yLy_{L}), slope (ySy_{S}), and term structure (yTy_{T}). A extreme scenario is followed by including the curvature component (yCy_{C}) to test four-dimensional control, yielding the vector 𝐲=(yL,yS,yC,yT)\mathbf{y}=(y_{L},y_{S},y_{C},y_{T}), which accounts for most of... | Figure 16 shows the distribution of absolute generation errors across all four features. The maximum absolute errors are approximately 5.7×10−45.7\times 10^{-4} for level, 4.6×10−34.6\times 10^{-3} for slope, 9.4×10−39.4\times 10^{-3} for curvature, and 2.0×10−32.0\times 10^{-3} for term-structure. The majority of feat... | Most absolute generation errors fall between 10−510^{-5} and 10−310^{-3} (equivalently, log10|e|∈[−5,−3]\log_{10}|e|\in[-5,-3]). The maximum absolute errors are approximately 5.5×10−45.5\times 10^{-4} for level, 8×10−48\times 10^{-4} for the slope over moneyness, and 5.4×10−45.4\times 10^{-4} for the term structure sl... | In the three-feature setting, we control y=(yL,yS,yT)y=(y_{L},y_{S},y_{T}) and keep the training hyperparameters of Experiment I, using β=5×10−8\beta=5\times 10^{-8}. | D |
=bc(t,Xtc)dt+ac(t,Xtc)dWt,\displaystyle=b_{c}(t,X_{t}^{c})\mathrm{d}t+a_{c}(t,X_{t}^{c})\mathrm{d}W_{t}, | Throughout this section, we will make the same assumptions that one makes when formally deriving the HJB equation. Specifically, we assume that an optimal control exists and that it is Markov. Under this assumption, for a fixed control cc, the controlled diffusion XcX^{c} is the solution of the following stochastic dif... | Now, we wish to find functions (p∗,c∗,λ∗,μ∗)(p^{*},c^{*},\lambda^{*},\mu^{*}) that together are a critical point of LL. A necessary condition for (p∗,c∗,λ∗,μ∗)(p^{*},c^{*},\lambda^{*},\mu^{*}) to be a critical point of LL is that the following functional derivatives are simultaneously equal to zero | where 𝒜c†\mathscr{A}_{c}^{\dagger} is the formal L2L^{2} adjoint of 𝒜c\mathscr{A}_{c}, which is given explicitly by | As solutions of SDEs are Markov processes, we can associated with XcX^{c} an infinitesimal generator of 𝒜c\mathscr{A}_{c}, which is defined as follows | D |
In contrast to exogenous transaction costs, endogenous transaction costs are those that are internal to the transaction and are directly related to the actions of the transacting parties. One of the most typical endogenous transaction costs is liquidity cost, resulting from liquidity risk, which is almost ubiquitous in... | Driven by the significance of exogenous and endogenous transaction costs, this paper analyzes how these factors influence an investor’s portfolio decisions under the utility-based approach, considering liquidity risks and transaction costs associated with asset trading. Based on the works of Pasricha et al. [34], we ad... | In this paper, we present a full study of the portfolio selection problem, considering both exogenous and endogenous transaction costs within the framework of utility maximization theory. Exogenous transaction costs are defined as proportional transaction costs, whereas endogenous transaction costs arise from liquidity... | In contrast to exogenous transaction costs, endogenous transaction costs are those that are internal to the transaction and are directly related to the actions of the transacting parties. One of the most typical endogenous transaction costs is liquidity cost, resulting from liquidity risk, which is almost ubiquitous in... | When transaction costs are taken into consideration, two important factors need to be considered in dynamic portfolio optimization. First, the portfolio cannot be continuously rebalanced, as it would lead to abnormally large trading costs. Therefore, the portfolio is hedged discretely in a non-infinitesimal fixed time ... | A |
Extending beyond traditional algorithms, Fish et al. (2024) run experiments with LLM-based pricing agents that autonomously collude, with outcomes sensitive to prompts. Keppo et al. (2025) further show that collusion is robust under some environments but fragile under others, depending on differences in model sophistic... | Our contribution is twofold. Existing work has largely established that algorithms can and do collude, both in simulations and in the field. Our addition is an analytical perspective that views cartel formation through a social learning lens, clarifying not only the mechanics of consensus but also the incentives of lea... | How cartels form (Coordination dynamics): In Section 6, we model collusion as a social learning process over discrete rounds of negotiation (DeGroot 1974), mediated by a shared pricing algorithm that all firms use.666Here, the “mediator” is simply the common pricing algorithm adopted across firms. It aggregates pooled ... | (ii) Cartel dynamics, where tacit coordination unfolds as a belief-averaging process (DeGroot 1974), mediated by shared pricing algorithms rather than direct communication, and leader–follower dynamics emerge as forceful agents systematically push consensus toward higher prices (Acemoglu et al. 2009); | Most of the existing literature has focused on showing that algorithmic agents can and do collude, both in simulations and in the field. Our contribution is complementary: we model cartel formation as a belief-averaging process (DeGroot 1974), enriched by forceful agents who pull outcomes toward their preferred prices ... | D |
SOFR+AXI is negatively correlated with the inflation rate, i.e., it increases during economic stress typically associated with declining prices. Overall, it exhibits correlations of sign and magnitude expected of a credit-sensitive rate and similar to those of LIBOR. However, the experience during the recent stress per... | The following section explains the key benefits of AXI and its methodology. Section 3 analyze its empirical properties via correlations with major credit indexes, core macroeconomic variables, and measures of economic uncertainty. To quantify AXI benefits, Section 4 analyzes profitability of hypothetical loans during t... | Table 2 reports correlations of SOFR+AXI and LIBOR with the Dodd-Frank Act Stress Testing (“DFAST”) 2025 core macroeconomic variables.212121For variable definitions see Federal | Reserve [2025a]. Following Greene [2017], the table reports significance of the coefficient in a univariate linear regression. These correlations should be viewed only as “suggestive,” as this quarterly dataset is small covering the period from Q2 2016 to Q3 2024.222222Both SOFR+AXI and LIBOR are transformed into quart... | Table 2: Correlations of quarterly changes in SOFR+AXI and LIBOR with the core macroeconomic variables | D |
(Xt,x,a,u,n,(Xt,x,a,u,n)′)=(Xt,x,a,u,n,λ(Xt,x,a,u,n,γt,a,u)).\left(X^{t,x,a,u,n},(X^{t,x,a,u,n})^{\prime}\right)=\left(X^{t,x,a,u,n},\lambda(X^{t,x,a,u,n},\gamma^{t,a,u})\right). | The value function vv defined in Theorem 13 is a viscosity solution of the HJB equation in the sense of Definition 9. Moreover, the map | Then, as n→+∞n\to+\infty, we recover the rough differential equation (40) in the sense of Theorem 5. | Suppose Assumption 3.1 holds. Then the value function defined in (43) is a viscosity solution of the rough HJB equation (47) with terminal condition (48), in the sense of Definition 9. | We now derive the rough Hamilton-Jacobi-Bellman (HJB) equation, which governs the value function in a rough path optimal control problem. In classical control theory, the HJB equation arises as a dynamic programming principle applied to the value function. When the system is driven by a rough path, however, the formula... | B |
To ensure a precise estimation of the necessary nuisance parameters in these estimators, we have developed a deep learning model comprising two main components: the NFR Net and the CNF Net. The NFR Net is highly effective in modeling complex, non-linear relationships, while the CNF Net excels in accurately estimating g... | We validate our methodology through extensive simulation studies and a real-world application using proprietary data from a major E-commerce platform. In simulations, the Dist-DML estimator consistently outperforms benchmark methods, including Distributional Direct Regression (Dist-DR) and Distributional Inverse Propen... | In line with previous studies, our results demonstrate a positive correlation between credit limits and consumer spending, underscoring the role of credit as a catalyst for consumption (Aydin,, 2022). Our analysis reveals a heterogeneous effect across different spending quantiles. In particular, as credit limits increa... | Through comprehensive numerical studies, we have demonstrated the effectiveness of our proposed Dist-DML estimator. In applying our approach to real-world data, we explored the causal effects of credit limit adjustments on consumer spending distributions. Our findings provide critical insights into consumer behavior, r... | The capacity of E-commerce platforms to set differentiated credit limits for individual users raises a critical research question: how does adjustment in credit limits influence consumer spending behaviors? To investigate this problem, we employ our approach by using data collected from a leading and large E-commerce p... | C |
∑i=1NGgi,t+∑j=1NP∑b=1Bϕj,b,t=0∀t:λt\displaystyle\sum^{N_{G}}_{i=1}g_{i,t}+\sum^{N_{P}}_{j=1}\sum^{B}_{b=1}\phi_{j,b,t}=0\quad\forall\quad t:\quad\lambda_{t}\hfill | Satellite models can run with a look-ahead window of arbitrary length to optimize for multi-day scheduling. For a given price, the result of the above optimization is the power dispatch timeseries pj,tp_{j,t} and not the demand or supply bid functions required by the central model. To generate bid functions consisting ... | The objective 1 minimizes the cost of electricity generation and the consumption costs of externally priced energy carriers (e.g., methane). Electricity, heat, and hydrogen are denoted by p,q,hp,q,h, respectively. Electricity generation and consumption must be balanced at all time steps 2. The prosumer electric power 3... | The objective maximizes social welfare defined as gross surplus minus generation costs 19. Equation 20 represents the market balance constraint, which consists of a set of bid blocks from each prosumer and a single bid block per generator. | Each prosumer maximizes revenues from electricity generated locally, or minimizes the consumption costs of energy carriers, subject to its technical constraints. The following problem is established for each prosumer. | C |
The proposed HyPV-LEAD addressed these limitations by combining window–horizon sequence modeling, hyperbolic embedding to encode hierarchical and hub–periphery structures, and PV sampling to emphasize peak/valley dynamics. Through this integration, HyPV-LEAD consistently outperformed all baselines, achieving an accurac... | The detailed ablation results are summarized in Table III and further illustrated in Fig. 1, which provide an intuitive comparison across model variants. | The proposed HyPV-LEAD addressed these limitations by combining window–horizon sequence modeling, hyperbolic embedding to encode hierarchical and hub–periphery structures, and PV sampling to emphasize peak/valley dynamics. Through this integration, HyPV-LEAD consistently outperformed all baselines, achieving an accurac... | In contrast, the proposed HyPV-LEAD, which integrates window–horizon based sequence modeling, hyperbolic embedding for hierarchical structure preservation, and PV sampling to emphasize volatility-driven segments, achieved the highest PR-AUC of 0.9624. This result demonstrates that the synergy among these three componen... | The overall comparison confirms that HyPV-LEAD delivers the most comprehensive improvements across metrics and establishes itself as an effective framework for anomaly detection in dynamic blockchain environments. The detailed performance of all models is presented in Table II. | D |
One of our key findings is that the leader’s equilibrium policy Π∗\Pi^{*} follows a Gaussian distribution. Moreover, its variance decreases as the volatility of the risky asset increases, holding other parameters fixed. In addition, the mean of the Gaussian distribution is independent of the randomization parameter λ0\... | In contrast to the pre-committed policy studied in Wang and Zhou (2020), the variance of our equilibrium policy does not necessarily decay over time. Instead, the constant variance we obtain is consistent with the equilibrium policy characterized in Dai et al. (2023). | One of our key findings is that the leader’s equilibrium policy Π∗\Pi^{*} follows a Gaussian distribution. Moreover, its variance decreases as the volatility of the risky asset increases, holding other parameters fixed. In addition, the mean of the Gaussian distribution is independent of the randomization parameter λ0\... | Now we consider the optimization problem of the leader. Formally, the leader’s problem is to optimize her objective functional subject to the follower’s best-response strategy characterized above. To evaluate the performance of a stochastic policy Π\Pi, we adopt the framework recently developed in the reinforcement lea... | We observe that the equilibrium strategy u2∗u_{2}^{*} in (3.9) is consistent with the equilibrium strategy derived in Theorem 3.2 of Huang and Sun (2023), where the intra-personal equilibrium strategy of NN investors with partial information under relative performance concerns is characterized. As explained in Huang an... | A |
The remainder of the paper is structured as follows: Section 2 describes the Hierarchical Risk Parity (HRP) methodology; Section 3 presents the dataset used in the study; Section 4 reports the empirical results; and Section 5 concludes with the main findings, limitations, and directions for future research. | where ww is the vector of asset weights, Σ\Sigma is the covariance matrix of asset returns, and μ\mu is the vector of expected returns. This formulation seeks the minimum variance portfolio for a given level of expected return μp\mu_{p}. The solution implies a closed-form expression: | Modern Portfolio Theory (MPT), developed by Harry Markowitz in 1952, introduced a formal and quantitative framework for portfolio construction by modeling the trade-off between expected return and risk [43]. In MPT, risk is represented by the variance (or standard deviation) of asset returns, while the expected return ... | This theory led to the formulation of the efficient frontier, a set of portfolios that offer the highest expected return for a given level of risk. Portfolio selection is then reduced to solving a constrained quadratic optimization problem, relying on the expected return vector and the covariance matrix of asset return... | Modern Portfolio Theory (MPT), pioneered by Harry Markowitz [29], laid the groundwork for the mathematical construction of portfolios, focusing on the relationship between expected return and risk [44]. However, in practice, MPT has certain limitations, such as high sensitivity to errors in parameter estimation [53], t... | B |
Optimal execution, the process of executing large orders in financial markets to minimize trading costs and market impact induced by large trades, is a cornerstone of modern financial research. The seminal work of AC (01) introduced a mean-variance framework that optimizes trading trajectories by considering both execu... | We first focus on analyzing the key points of this paper, the time-dependent fee schedule. Figure 1(a) displays the fee ctlc_{t}^{l} charged in the lit market. Within estimation error it is flat, taking the constant value 0.010.01, which is the upper bound of transaction fees, over the entire trading horizon [0,1][0,1]... | Optimal liquidation has been well investigated in the literature and the context of a classical venue, named lit pool, as emphasized above. | Note that intuitively, the definition of QQ, 𝒱\mathcal{V} and ℒ\mathcal{L} avoid QQ to become negative. Suppose at time tt the order in at least one dark pool is executed, the instantaneous trading volume in the lit market is zero, then the trader will continuously update the trading rate regarding the updated invento... | Figure 4(a) to 4(c) depict the trading rate ν^t\hat{\nu}_{t} in the lit market. For both regulated scenarios, ν^t\hat{\nu}_{t} is strictly increasing, starting from approximately −1.75-1.75 at t=0t=0 and ending near −0.20-0.20 at t=1t=1. In economic terms the big trader initially traders conducted rapid trading in the ... | B |
The theoretical analysis delivers two main results. First, with risk-neutral buyers, accurate certification is always more profitable in equilibrium than any equilibrium with noisy certification. The informativeness loss dominates the differentiation gain. Second, with loss-averse buyers, this ranking can reverse. Loss... | Whether noisy certification raises profitability is a priori ambiguous. Outcomes depend on loss aversion, certification accuracy and cost, and the distribution of quality. Our theoretical analysis isolates the mechanisms and the conditions under which each dominates but does not deliver a universal ranking, so we compl... | The theoretical analysis delivers two main results. First, with risk-neutral buyers, accurate certification is always more profitable in equilibrium than any equilibrium with noisy certification. The informativeness loss dominates the differentiation gain. Second, with loss-averse buyers, this ranking can reverse. Loss... | Within the limited literature on certification under oligopoly, our paper is related to De and Nabar (1991), Bottega and Freitas (2019), and Zhang and Li (2020). In De and Nabar’s framework, the market is perfectly competitive and oligopolistic sellers are price-takers. Our setting is different as sellers can strategic... | Since differences in paid certification costs cannot fully explain the observed profitability ranking, we compare the competitiveness of the environments. As discussed in the theoretical section, noisy certification introduces endogenous product differentiation by adding randomness to buyers’ WTP, which can reduce comp... | A |
For convenience, the matrices ℬ,𝒞\mathcal{B},\mathcal{C} and vectors 𝐑𝐇𝐒(n),𝐑𝐑𝐇𝐒(n),𝐛,𝐜{\bf{RHS}}^{(n)},{\bf{RRHS}}^{(n)},\bf{b},\bf{c} are presented in the Appendix A. To efficiently compute the inverses of tridiagonal matrices ℬ\mathcal{B} and 𝒞\mathcal{C}, we employ LU decomposition which decomposes the m... | Since the pricing of American options we study is a highly nonlinear problem, it is difficult to find its analytical solution. In order to validate our numerical scheme, we first trace back to the European option pricing without transaction costs, whose closed-form solution has been presented in [46]. From Table 1, we ... | In this paper, the pricing problem of American options is studied when both exogenous and endogenous transaction costs are taken into consideration. While endogenous transaction costs here are referred to as liquidity risks, which are modeled with an Ornstein-Uhlenbeck process, exogenous transaction costs are associate... | Table 1: European option prices obtained by the closed-form solution in [46] and our numerical scheme with κ=0\kappa=0. The specific space and time steps are in the form of (NS,NL,NT)(N_{S},N_{L},N_{T}). | Table 5 displays that the relative computational error between the results from the ADI method and those from the fully explicit method is still less than 0.61%0.61\% when valuing American options with transaction costs, which demonstrates that our scheme is correct. Moreover, compared to the fully explicit scheme, the... | A |
In the trajectory forecasting approach, researchers predict an entire sequence of future prices—referred to as a price trajectory—to capture the expected development of prices over time. | Within the RW forecasting setup, there are two important moments in time: the forecasting time, which is the moment the forecast is made, and the forecasted period, which is the period for which a forecast is made. | However, the complete trajectory is forecasted at once, without incorporating new market information that may become available during the forecasting horizon. | Our results confirm that incorporating trading information from other CID market products—especially those traded simultaneously with the target product—improves price forecasting accuracy. | Since the delivery periods of neighboring products differ from that of the forecasted product, trading in some neighboring products may cease during the forecasting period, rendering trade data for these products unavailable. | B |
𝐜t\mathbf{c}_{t}Cell×\timesforgetforgetht−1lh_{t-1}^{l}htl−1h_{t}^{l-1}inputinputht−1lh_{t-1}^{l}htl−1h_{t}^{l-1}outputoutputht−1lh_{t-1}^{l}htl−1h_{t}^{l-1}tanh×\times×\times𝐡tl{\mathbf{h}^{l}_{t}}ht−1lh_{t-1}^{l}htl−1h_{t}^{l-1} | The non-linear transformation, σ\sigma, triggered in each gate is by the rectified linear unit (ReLU) activation function, i.e. σ=ReLU(x)=max(x,0)\sigma=ReLU(x)=max(x,0) 666Other widely used nonlinear activation functions include the sigmoid (also referred to as the logistic function) and the tanh (also referred... | The following properties of the pseudo-SNAP model explain why our model matters: 1) flexible functional form for each pricing element and high dimensional information sets: Each pricing element is able to be independently represented by a nonlinear and flexible network function, such as convolutional neural network (CN... | Note: This figure illustrates the LSTM recurrent network memory cell. The cells are linked to each other and three gates, a forget gate, input gate, and output gate regulate information flow in each cell. All the gating units have a nonlinear activation function, we use rectified linear unit (ReLU) activation function. | The key to LSTM is the memory cell that has the function to add or remove information from previous and current states through three gates, forget gate, input gate, and output gate (Graves, 2014). Figure 2 summarizes the detailed architecture of the memory cell. | C |
∑t=0TP(D=t)(u(c0)+…+u(ct))=∑t=0T(1−m)tu(ct).\displaystyle\sum_{t=0}^{T}P(D=t)(u(c_{0})+...+u(c_{t}))=\sum_{t=0}^{T}(1-m)^{t}u(c_{t}). | We find that the discount rate is no longer constant because the finite-time human extinction violates the “perpetual youth” assumption. However, for large tt it converges to a constant. Specifically, the long-run discount factor is equal to (1−M)(1−m)(1+b)<1(1-M)(1-m)(1+b)<1. This means—exactly like in the case of t... | Now the dynasty’s discount rate is the difference between the aggregate extinction risk MM, and the population growth rate (the birth rate bb minus the individual mortality rate mm), taken to the power θ\theta. If the dynasty size is constant over time, the discount rate is again equal to the extinction hazard rate MM.... | It follows that the dynasty’s discount rate is the sum of the individual death rate mm and the aggregate extinction risk MM, minus the birth rate bb. Specifically, if the dynasty size is constant over time, so that n=0n=0 (i.e., (1+b)(1−m)=1(1+b)(1-m)=1), then the discount rate is equal to the extinction hazard rate M... | This implies that when the extinction date is known, the discount rate reflects only the individual death rate mm—i.e., only the uncertainty in one’s lifespan, not its expected length. Indeed, the discount rate is exactly the same regardless of the extinction date TT, which means it disregards the prospects of future e... | D |
Our findings present several compelling avenues for future research, building upon the foundational framework we have established. An immediate next step is to examine whether the trends observed in the ETH-USDC 5bps and 30bps pools hold for other token pairs. While a full analysis of every possible pair is beyond the ... | A key finding is the stability of the empirical PCA basis for the 5 bps pools (Ethereum ETH-USDC and Arbitrum ARB-USDC). This low-rank structure is highly stable across rolling time windows and also aligns consistently with a low-order Legendre polynomial basis. This alignment is a critical discovery, as it provides a ... | From a similar perspective, it is worth investigating if the basis stability, particularly the alignment with the Legendre basis, holds for a wider array of dominant pools, such as those involving memecoins, stablecoin-stablecoin pairs, or the BTC-ETH pair. We anticipate that the stylized facts established in this stud... | Basis Alignment: For the 5 bps pools (ARB-USDC and ETH-USDC), the empirical PCA bases strongly align with the basis of Legendre polynomials. This consistency across time windows allows for the use of a parsimonious, fixed-factor model. The distributional facts mentioned above also hold for the Legendre coefficient seri... | Our findings present several compelling avenues for future research, building upon the foundational framework we have established. An immediate next step is to examine whether the trends observed in the ETH-USDC 5bps and 30bps pools hold for other token pairs. While a full analysis of every possible pair is beyond the ... | B |
This work extends and complements our previous theoretical paper on the subtle interplay between impact, order flow and volatility [1]. In that work, most of our predictions turned out to be in rather remarkable agreement with empirical observations, despite the simplifying mathematical approximations that we had to ma... | The framework we introduce here allows one to justify further our proposal using purely simulated data. Although we have not yet been able to compute exactly the impact of proxy metaorders within our model, we believe that our numerical results are convincing enough to believe that the procedure proposed in Ref. [2] is... | Furthermore, we were able to construct proxy metaorders from simulated order flow that reproduce the square-root law of market impact — a law that has long been, and in some circles still is, attributed to information revelation; see e.g. [15, 16, 17]. Our model, on the other hand, makes the assumption that impact is p... | This work extends and complements our previous theoretical paper on the subtle interplay between impact, order flow and volatility [1]. In that work, most of our predictions turned out to be in rather remarkable agreement with empirical observations, despite the simplifying mathematical approximations that we had to ma... | Finally, to understand price formation from order flow, we introduced a generalized propagator model. This instrument is crafted to incorporate the three main stylized characteristics of the impact of metaorders (see [3, 6] and refs therein): (i) impact grows on average as the square-root of the number of child orders ... | B |
Each column represents an order book state. Moving along the width dimension from column 1 to TT moves forward in time. | LOBs are the dominant microstructure for trading in modern electronic financial markets (Gould et al., 2013). They operate in continuous time, allowing participants to submit buy and sell orders at any price and size. It is a continuous double auction, such that when a buy and sell order cross — the bid, or buy price, ... | Figure 1. An example of the forward diffusion process. The top row shows the original data in the image format, with prices on the left and sizes on the right. Each column represents a snapshot of the order book at each timestep. The bottom row shows a fully noised input to the model, which would be fed in at inference... | Each row represents a price level. Moving along the height dimension from 1 to 2n2n increases in price, i.e. starting from the lowest, least competitive bid price, we move towards the highest, most competitive bid price. The lowest, most competitive ask price is directly above the best bid, and the highest, least comp... | The authors of (Li et al., 2025) use the order batch model to model longer term dependencies, as discussed in the generative section above, encoding order batches as images in a unique way to provide a summary of the number of bid, ask and cancel orders at each price and volume level. However, as this method serves to ... | C |
In case (a), a low learning elasticity leads to a unique, globally stable interior steady state. In case (b), the interior steady state becomes unstable, and the dynamics exhibit threshold behavior: if the initial productivity of sector 1 is below the unstable steady state, the planner shifts employment fully to sector... | Figure 6(a) shows a case with a moderate endorphins effect, resulting in a unique and globally stable interior steady state at a low fitness level. In Figure 6(b), a stronger endorphins effect gives rise to two interior locally stable steady states — one at a low fitness level, which we call a "couch potato" steady sta... | While Theorem 1 provides a general method for identifying and classifying the stability of interior steady states using the locator function, this section highlights important cases in which additional analysis yields sharper insights. We begin by showing that when the locator function crosses zero only once and from a... | We also consider what is arguably the typical non-convex case: a locator function that is inverted U-shaped with two interior roots. In this setting, we show that the only possible locally stable steady state in the interior of the state space is the highest root. Moreover, we show that this is indeed the case if the p... | The locator function in the lower panels successfully captures these dynamics. In case (a), it has a single interior root with negative slope, corresponding to a globally stable interior steady state, as implied by Proposition 4.1 and Lemma 5.3. In case (b), the locator function has a single interior root with positive... | D |
The role of iridium in PEMEL expansion has been explored through various projections of demand, supply constraints, and technological progress. One of the earliest studies, by Smolinka et al. [8], focused on Germany and concluded that, without major reductions in catalyst loading, national iridium demand could exceed g... | The role of iridium in PEMEL expansion has been explored through various projections of demand, supply constraints, and technological progress. One of the earliest studies, by Smolinka et al. [8], focused on Germany and concluded that, without major reductions in catalyst loading, national iridium demand could exceed g... | The first parameter for estimating iridium demand is the annual rate of PEMEL capacity installation. Two scenarios are used. The first is based on real and planned PEMEL projects compiled in the IEA database, which covers developments till 2030 [24] and is referred to as Business-As-Usual scenario (BAU). Capacity growt... | This study addresses key research gaps by developing a recursive model that estimates iridium demand from both initial PEMEL installations and end-of-life replacements, incorporating recycling effects. It also introduces refined supply scenarios by accounting for competing market demands and iridium price trends. Deman... | Figure 4: The conservative IEA-NZE scenario: Projected iridium demands with an average lifetime τ=10\tau=10 years and recycling efficiency of 97% by 2035 and corresponding supply projections. a) Demand projection, associated recycling curve and supply projections. b) Supply to demand gaps under strong supply. c) Supply... | C |
For each t≤Tt\leq T, VtV_{t} is centered Gaussian conditional on {Hu}u≤t\{H_{u}\}_{u\leq t} with variance | 𝔼[(Vt−Vs)2∣{Hu}u≤t]=∫0s(K(t,u)−K(s,u))2𝑑u+∫stK(t,u)2𝑑u.\mathbb{E}\big{[}(V_{t}-V_{s})^{2}\mid\{H_{u}\}_{u\leq t}\big{]}=\int_{0}^{s}\!\!\big{(}K(t,u)-K(s,u)\big{)}^{2}du\;+\;\int_{s}^{t}\!\!K(t,u)^{2}du. | At=∫0tK(t,u)2𝑑u<∞A_{t}=\int_{0}^{t}K(t,u)^{2}du<\infty. The process VV admits a continuous modification. | σt=V0exp(νVt−ν22At),At:=∫0tK(t,u)2𝑑u,\sigma_{t}=\sqrt{V_{0}}\,\exp\!\Big{(}\nu V_{t}-\tfrac{\nu^{2}}{2}A_{t}\Big{)},\qquad A_{t}:=\int_{0}^{t}K(t,u)^{2}\,du, | ∫0tK(t,u)2𝑑u≤1cΓ2∫0t(t−u) 2ε−1𝑑u=t2ε2εcΓ2for all t∈(0,T],a.s.\int_{0}^{t}K(t,u)^{2}\,du\;\leq\;\frac{1}{c_{\Gamma}^{2}}\int_{0}^{t}(t-u)^{\,2\varepsilon-1}\,du\;=\;\frac{t^{2\varepsilon}}{2\varepsilon\,c_{\Gamma}^{2}}\qquad\text{for all }t\in(0,T],\ \text{a.s.} | B |
As shown in Fig. 7(a) for American Put prices, the neural network predicted prices align well with the ground truth prices, with all of the predicted prices distributing nicely around the diagonal line of perfect matches. Fig. 7(b) shows a detailed breakdown of the distribution of prediction errors Err=V′−VErr=V^{\pr... | where c(K,T)c(K,T) and p(K,T)p(K,T) are the price of the call and put option, respectively, at strike KK and time to maturity TT. S(0)S(0) is the spot value of the underlying, rr is the risk-free rate, N(⋅)N(\cdot) is the standard normal cumulative distribution function. For any strike KK and time to maturity TT wh... | Figure 8: Benchmark the performance of the neural network pricer for Asian call and Asian put. (a) Comparison between the ground truth price VV and neural network-predicted price V′V^{\prime} for Asian call. (b) Distribution of Err=V′−VErr=V^{\prime}-V of all test pricing data in the log moneyness kk and time to matu... | Moreover, Fig. 8(a) and (c) show the comparison between the neural network predicted prices versus the ground truth prices for arithmetic Asian call and put options. For both types of Asian options, the prices again agree very well. Similarly, Fig. 8(b) and (d) show the error in the kk and TT plane. Similar to the Amer... | As shown in Fig. 7(a) for American Put prices, the neural network predicted prices align well with the ground truth prices, with all of the predicted prices distributing nicely around the diagonal line of perfect matches. Fig. 7(b) shows a detailed breakdown of the distribution of prediction errors Err=V′−VErr=V^{\pr... | C |
Rationale: The most subtle and important type of structural break is concept drift, when the distributional relationship between the features and the outcome has fundamentally changed. For example, an indicator that is previously important in nowcasting market capitulation is now insignificant for a different time peri... | Implementation: We perform a SHAP stability analysis. We split the hold-out test set chronologically into 2 halves, and assess global SHAP feature importance bar plots for the first half and for the second half of the test dataset. Any considerable change in the rank or magnitude of importance of features, between thes... | Notes: This figure assesses the stability of the model’s feature interpretations over time to test for concept drift. The panels display the mean absolute SHAP values for the top 20 features from the primary SVM model, calculated independently for two chronological sub-periods of the hold-out test set (July 2023 - June... | Results: As seen in Figure 9, the SHAP importance plots are very much in line with each other across both halves of the hold-out. The highest ranked features in the first half (Panel (a)) remained the highest ranked features in the second half (Panel (b)), and their contributions are similar. In fact, the most SHAP-sig... | To move beyond global importance and explore nonlinear relationships, we examine SHAP dependence plots. These plots show how a feature’s marginal contribution to the prediction (its SHAP value) changes across the range of its values. Figure 6 illustrates these relationships for our most influential predictors, revealin... | A |
The adaptedness (or bi-causality) imposed on couplings modifies the Wasserstein distance to ensure robustness of a large class of “well-defined" dynamic optimization problems, ranging from optimal stopping and utility maximization to dynamic risk minimization and dynamic hedging; see | Robustness (Bartl and Wiesel (2023); Pflug and Pichler (2014); Acciaio et al. (2020, 2024a)). Let μ,ν∈𝒫2(ℝdT)\mu,\nu\in\mathcal{P}_{2}(\mathbb{R}^{dT}) and v:𝒫2(ℝdT)→ℝv\colon\mathcal{P}_{2}(\mathbb{R}^{dT})\to\mathbb{R}, where v(μ)v(\mu) denotes the optimal value of a “well-defined” dynamic optimization problem... | A large class of generative AI models for financial time series is trained for decision-making applications, including dynamic hedging, optimal stopping, utility maximization, and reinforcement learning. Notably, these problems are not continuous with respect to widely-used distances, such as the Maximum Mean Discrepan... | From the perspective of nested disintegration, Pflug-Pichler define the adapted Wasserstein distance as nested distance in Pflug and Pichler (2014) and establish an alternative representation of 𝒜𝒲2(⋅,⋅)\mathcal{AW}_{2}(\cdot,\cdot) through the dynamic programming principle; see the proof of Proposition 1 in Append... | Bion-Nadal (2009); Glanzer et al. (2019); Bartl and Wiesel (2023); Pflug and Pichler (2014); Acciaio et al. (2020, 2024a). We therefore present only the general statement here, while providing illustrative examples in the appendix and referring the reader to the cited references for further details. | D |
for the larger average difference in means (LABEL:tab:table-cultgroup-mean-rankdiffs-LSGWP-vs-CLGWP-standalone | are drawn from global (Normal) distributions:222Formally, 𝜷j≡(βj1,…,βjK)⊤.\boldsymbol{\beta}_{j}\equiv(\beta_{j1},\ldots,\beta_{jK})^{\top}. | yi∼𝒩(αj+𝐗i⊤𝜷j,σ)y_{i}\sim\mathcal{N}\big{(}\alpha_{j}+\mathbf{X}_{i}^{\top}\boldsymbol{\beta}_{j},\,\sigma\big{)} | income of ∼0.34\sim 0.34 on the eleven-point scale. the negative impact of being unemployed is enormous — | ∼𝒩(μβk,σβk),k=1,…,K,\displaystyle\sim\mathcal{N}(\mu_{\beta_{k}},\,\sigma_{\beta_{k}}),\quad k=1,\ldots,K, | B |
All these functions are assumed to exist and to be computable analytically or at least numerically. The first result is the following: | Assuming that all PPs are probability density function and that S1(t)S_{1}(t) and S2(t)S_{2}(t) are absolutely continuous, then the last particle survival function S1(t)S^{1}(t) can be computed using the following formula: | the marginal survival probability Si(t)S_{i}(t) and the marginal first passage time density (if it exists) Fi(t)F_{i}(t) for t<τmt<\tau_{m}. | {(s0,0),(s1,τ1),…,(sn−1,τn1),(s,t)}\{(s_{0},0),(s_{1},\tau_{1}),\ldots,(s_{n-1},\tau_{n_{1}}),(s,t)\}, where s0s_{0} is the initial state where no coordinate has been killed yet. Then the contribution to the nthn\textsuperscript{th} survival function given by gg is defined by the following integral: | If we assume a killing barrier at M>0M>0, the survival function describing the probability that both coordinates are below this barriers is | A |
The vector 𝐰\mathbf{w} represents the portfolio weights allocated to four distinct trading experts over the next trading window: the trend expert executes trades based on momentum signals, the reversal expert specializes in mean-reversion strategies using contrarian signals, the breakout expert capitalizes on event-dr... | Automated financial trading systems serve as essential tools for efficient capital allocation and risk management in financial markets. However, financial markets are inherently non-stationary, frequently shifting due to policy changes, macroeconomic cycles, and black-swan events. This dynamic complexity poses fundamen... | These consistent results indicate that candlestick chart patterns and temporal features are not merely redundant but complementary: visual modality provides critical pattern signals (such as candlestick combinations) that require quantitative validation from temporal data in the price-volume dimension, with their syner... | (iii) Four heterogeneous domain-specialised experts run in parallel, each optimising a single trading archetype. Their weighted outputs are finally aggregated into an executable portfolio. This separation of macro regime perception from micro strategy execution balances diversity, interpretability, and adaptability to ... | All experts receive the same multimodal observation sts_{t}, yet their parameters are optimised separately so that their decision logics remain diverse. | D |
The high-volatility market, which involves over $40 trillion in market capitalization as well as significantly higher risk and profit opportunity, has attracted the attention of innumerable investors around the world. The straddle option, which is designed for scenarios where a trader anticipates significant market vol... | Building on these advancements, our model uses self-attention mechanisms for time series data to quickly capture the latest market information and optimize asset weights accordingly. We also integrate channel attention mechanisms to balance short-term adjustments with a strategic view of long-term trends for higher lon... | As shown in Figure 4 and Table II, the performances of the three model variants are inferior to that of the complete Transformer-DDQN model. For NoRes-Transformer-DDQN, removing resistance information causes the model to focus only on current volatility and neglect key regions of recent long-short battles, increasing t... | To hunt optimal timing for the straddle option to open and close the position, we face the following two major challenges: i) How to hunt the optimal timing to trade and adopt different strategies at various points of market volatility; ii) How to enable the model to understand long-term trends while focusing on short-... | ii) Channel attention mechanism allows the model to balance short-term reactive adjustments with a strategic understanding of long-term trends, thereby optimizing for long-term gains. | C |
Parametric option pricing models are workhorses of contemporary academic research and industry practice. At a high level, the selection of a particular model is to specify the stochastic process for the underlying asset’s price as well as for any other relevant state variables, with the parameters of these processes of... | Even if the data generating process is not known, the minimization problem is still conceptually valid as a means of obtaining the best fitting parameters. The "best" model obtained in this context is the one that minimizes the objective function. A potential issue is that while the resulting model might match option p... | Broadly speaking, the calibration of an option pricing model involves minimizing some function of the model fitting error to search for the a set of parameters which minimizes this error, thereby producing the best fit. Generally, the function being minimized is a weighted sum of squared errors, either in terms of pric... | The calibration of option pricing models is a routine procedure across both industry practice and academic research. This article offers a calibration procedure based on an objective function which jointly considers errors in the implied volatility surface and errors in the term structure of variance. My objective func... | One might wonder of what additional benefit can be had by using the information from the variance swap or VIX term structures; after all, the term structure of VIX is itself identified from option prices with the variance swap rate and VIX being highly correlated. However, as I explore further in section 4, fitting the... | B |
If the DSO can accurately forecast each customer’s total energy demand over an optimization horizon, they can use the results to calculate α\alpha coefficients in the LRP. For example, the DSO use an optimal powerflow analysis to determine the optimal load profile for controllable loads on a distribution feeder. This “... | The optimal-α\alpha LRP shares some similarities to the Dynamic Power Tariff (DPT) proposed in Huang et al. [24], though the prices are constructed in different ways. Huang et al. build on their quadratic DLMP pricing work [23] by proposing a quadratic power tariff with prices determined through an iterative process be... | To address this problem, we separate the calculation of quadratic costs into a volumetric energy price, similar to conventional day-head prices, and a pricing component that is linearly dependent on quantity. This composite price, referred to in this paper as Load-Responsive Pricing (LRP), is then communicated to custo... | In the optimal-α\alpha LRP, we separate the calculation of αt\alpha_{t} prices to occur after target load profiles and βt\beta_{t} energy prices are calculated. This decoupled approach allows the DSO to use any OPF strategy they wish to identify the desired load profile. At the same time, βt\beta_{t} prices can be dete... | Although researchers have explored many different price-signal calculation mechanisms, each of the price signals in the papers referenced above lead to a total customer cost that is linear in quantity. This type of problem can be solved by each customer via a linear program. However, in [23], Huang et. al. explore quad... | A |
To extend this to rare-event detection, we incorporate volatility diagnostics based on recursive Fibonacci windows (21,34,55,8921,34,55,89). | and compute the Mahalanobis distance of the simulated section points relative to the observed section cloud: | D2(ziθ)=(ziθ−μ)⊤Σ−1(ziθ−μ),D^{2}(z_{i}^{\theta})=(z_{i}^{\theta}-\mu)^{\top}\Sigma^{-1}(z_{i}^{\theta}-\mu), | By comparing both models on the same system, we identify how diagnostic weighting alters posterior structure. | For a simulated trajectory under proposal θ\theta, we compute both (i) the correlation integral Cθ(r)C_{\theta}(r) | D |
Accuracy Reward Prediction accuracy is commonly used to evaluate the performance of reinforcement learning models and construct loss functions. Specifically, it measures the consistency between the model’s predictions and the ground-truth outcomes (rise/fall) of each sample. | In this work, we propose the FinZero model, as illustrated in Figure˜1, which fine-tunes 3B-parameter multimodal large model via the UARPO method in the FVLDB dataset, which enables MLM to explicitly account for prediction uncertainty. Comparative experiments with GPT-4 show a 13.48% improvement in prediction accuracy ... | As shown in Figure˜3, the model’s rewards continuously increase during the UARPO fine-tuning process: the format reward and completion length reward rise rapidly in the early stage of training and then stabilize, while the accuracy reward also increases steadily with training; meanwhile, the loss value decreases consis... | Confidence Score Prior works ([35, 36]) have explored the feasibility and methods for large models to learn task uncertainty. Given the high uncertainty inherent in financial decision-making—where uncertainty analysis is critical for model development and real-world use—we integrate model reasoning uncertainty into rei... | Completion Length Reward Previous works have found that text length expansion occurs in large model RL reasoning, which is helpful for improving reasoning time and enabling complex reasoning. Therefore, we provide this type of reward. Specifically, when the text reasoning length is no more than 200 tokens, a gradually ... | D |
The remainder of this paper is organized as follows: Section 2 provides the institutional background, while Section 3 discusses the data and presents the reduced-form evidence. Section 4 outlines the model, followed by the estimation approach and results in Section 5. Section 6 presents the counterfactual analysis, and... | China, like many other developing countries, has a less established system for providing primary healthcare services. On one hand, China faces growing burden of aging society and chronic diseases. On the other hand, China’s public hospitals are only major providers for most medical services and healthcare needs. Conseq... | Specifically, township health centers in rural areas are tasked with screening, early detection, and health management of hypertension and diabetes. Long-term care for chronic diseases is also provided through outpatient ambulatory services at public hospitals, with these services typically covered by government health... | Chronic diseases stand as leading causes of death and major contributors to healthcare costs worldwide. The escalating prevalence of chronic diseases not only increases healthcare expenditures but also poses a significant challenge to global efforts in reducing poverty and enhancing health equity. Disadvantaged populat... | Chronic diseases are considered as the leading causes of death and the major drivers of healthcare costs across countries. The World Health Organization (WHO) recommends using primary healthcare for early detection and timely treatment as the critical approach to managing chronic diseases.333See, for example, the Fact ... | D |
README.md exists but content is empty.
- Downloads last month
- 11