Tune DNN hyperparameters for binary classification using Optuna TPE search.
npx skills add https://github.com/aliyun/qwen-dianjin --skill dnn-tuning
Please help me install this Agent Skill. Skill: dnn-tuning Source: https://github.com/aliyun/qwen-dianjin/tree/main/DianJin-SKILLS/financial-engineering-expert/dnn-tuning Command: npx skills add https://github.com/aliyun/qwen-dianjin --skill dnn-tuning
This Skill helps you systematically tune a deep neural network (DNN) for binary classification by diagnosing overfitting/underfitting and then running Optuna-based search over network architecture and training hyperparameters to improve validation performance and generalization on OOT.
Ask the agent to run DNN tuning with your data path and target column in interactive mode, assuming you have already trained a baseline model with dnn-modeling.
High-intent search queries and answers about installing and using this skill.
Diagnosis-driven hyperparameter tuning analyzes train and validation performance gaps to identify overfitting or underfitting, then dynamically constrains the Optuna TPE Bayesian search space to target the specific diagnosed issue for improved AUC and KS.
Diagnose DNN overfitting by evaluating train and validation performance gaps across classic three-split data setups, then adjusting dropout, weight decay, and network architecture based on the diagnosis to improve OOT generalization.
Yes, Optuna hyperparameter tuning requires a pre-trained DNN baseline model created via dnn-modeling, providing the initial train/val/OOT split structure needed to diagnose behavior and guide the constrained search space.
Yes, you can automate multi-round DNN tuning using AUTO mode, which iterates automatically until convergence criteria are met, while interactive mode pauses each round for user feedback.
Bayesian search optimizes DNN architecture parameters including n_layers, layer_width, and dropout, alongside training parameters like learning_rate, weight_decay, and batch_size to maximize validation performance.
aliyunTrain PyTorch MLP binary classifiers and generate AUC/KS/BCR/Lift evaluation reports.
aliyunConvert features into WoE bins and fit Logistic Regression for risk scorecards.
aliyunTrain segmented XGBoost sub-models and fuse them via out-of-fold stacking.
h4vzzExplore hyperparameter configurations using grid, random, Bayesian, or Hyperband strategies.
