trustformers_debug/model_diagnostics/

auto_debug.rs

//! Auto-debugging and automated recommendation system.
//!
//! This module provides intelligent debugging capabilities that automatically
//! analyze model behavior, identify potential issues, and generate actionable
//! recommendations for optimization and problem resolution.

use anyhow::Result;
use std::collections::{HashMap, VecDeque};

use super::types::{
    ConvergenceStatus, LayerActivationStats, ModelPerformanceMetrics, TrainingDynamics,
};

/// Auto-debugging system for intelligent model analysis.
#[derive(Debug)]
pub struct AutoDebugger {
    /// Debugging configuration
    config: AutoDebugConfig,
    /// Historical performance data for analysis
    performance_history: VecDeque<ModelPerformanceMetrics>,
    /// Layer statistics history
    layer_history: HashMap<String, VecDeque<LayerActivationStats>>,
    /// Training dynamics history
    dynamics_history: VecDeque<TrainingDynamics>,
    /// Known issue patterns and solutions
    #[allow(dead_code)]
    issue_patterns: IssuePatternDatabase,
    /// Current debugging session state
    session_state: DebuggingSession,
}

/// Configuration for auto-debugging system.
#[derive(Debug, Clone)]
pub struct AutoDebugConfig {
    /// Maximum history size for analysis
    pub max_history_size: usize,
    /// Minimum samples required for pattern detection
    pub min_samples_for_analysis: usize,
    /// Confidence threshold for recommendations
    pub recommendation_confidence_threshold: f64,
    /// Enable advanced pattern recognition
    pub enable_advanced_patterns: bool,
    /// Enable hyperparameter suggestions
    pub enable_hyperparameter_suggestions: bool,
    /// Enable architectural recommendations
    pub enable_architectural_recommendations: bool,
}

/// Current debugging session state.
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
pub struct DebuggingSession {
    /// Session start time
    pub session_start: chrono::DateTime<chrono::Utc>,
    /// Issues identified in current session
    pub identified_issues: Vec<IdentifiedIssue>,
    /// Recommendations generated
    pub recommendations: Vec<DebuggingRecommendation>,
    /// Session statistics
    pub session_stats: SessionStatistics,
}

/// An identified issue with diagnostic information.
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
pub struct IdentifiedIssue {
    /// Issue category
    pub category: IssueCategory,
    /// Issue description
    pub description: String,
    /// Severity level
    pub severity: IssueSeverity,
    /// Confidence in identification
    pub confidence: f64,
    /// Evidence supporting the identification
    pub evidence: Vec<String>,
    /// Potential causes
    pub potential_causes: Vec<String>,
    /// When issue was identified
    pub identified_at: chrono::DateTime<chrono::Utc>,
}

/// Categories of issues that can be identified.
#[derive(Debug, Clone, PartialEq, Eq, Hash, serde::Serialize, serde::Deserialize)]
pub enum IssueCategory {
    /// Learning rate related issues
    LearningRate,
    /// Gradient flow problems
    GradientFlow,
    /// Overfitting issues
    Overfitting,
    /// Underfitting issues
    Underfitting,
    /// Data quality problems
    DataQuality,
    /// Architecture inefficiencies
    Architecture,
    /// Memory management issues
    Memory,
    /// Convergence problems
    Convergence,
    /// Numerical stability issues
    NumericalStability,
}

/// Severity levels for identified issues.
#[derive(Debug, Clone, PartialEq, PartialOrd, serde::Serialize, serde::Deserialize)]
pub enum IssueSeverity {
    /// Minor issue with low impact
    Minor,
    /// Moderate issue affecting performance
    Moderate,
    /// Major issue requiring attention
    Major,
    /// Critical issue requiring immediate action
    Critical,
}

/// Auto-generated debugging recommendation.
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
pub struct DebuggingRecommendation {
    /// Recommendation category
    pub category: RecommendationCategory,
    /// Recommendation title
    pub title: String,
    /// Detailed description
    pub description: String,
    /// Specific actions to take
    pub actions: Vec<String>,
    /// Expected impact
    pub expected_impact: String,
    /// Confidence in recommendation
    pub confidence: f64,
    /// Priority level
    pub priority: AutoDebugRecommendationPriority,
    /// Relevant hyperparameters to adjust
    pub hyperparameter_suggestions: Vec<HyperparameterSuggestion>,
}

/// Categories of recommendations.
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
pub enum RecommendationCategory {
    /// Hyperparameter adjustments
    HyperparameterTuning,
    /// Architectural changes
    ArchitecturalModification,
    /// Data preprocessing recommendations
    DataPreprocessing,
    /// Training strategy changes
    TrainingStrategy,
    /// Debugging and monitoring
    DebuggingAndMonitoring,
    /// Resource optimization
    ResourceOptimization,
}

/// Priority levels for recommendations.
#[derive(Debug, Clone, PartialEq, PartialOrd, serde::Serialize, serde::Deserialize)]
pub enum AutoDebugRecommendationPriority {
    /// Low priority suggestion
    Low,
    /// Medium priority recommendation
    Medium,
    /// High priority action needed
    High,
    /// Urgent action required
    Urgent,
}

/// Hyperparameter adjustment suggestion.
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
pub struct HyperparameterSuggestion {
    /// Parameter name
    pub parameter_name: String,
    /// Current value (if known)
    pub current_value: Option<f64>,
    /// Suggested value
    pub suggested_value: f64,
    /// Adjustment reasoning
    pub reasoning: String,
    /// Expected effect
    pub expected_effect: String,
}

/// Database of known issue patterns and solutions.
#[derive(Debug, Clone)]
pub struct IssuePatternDatabase {
    /// Learning rate patterns
    pub learning_rate_patterns: Vec<IssuePattern>,
    /// Gradient patterns
    pub gradient_patterns: Vec<IssuePattern>,
    /// Convergence patterns
    pub convergence_patterns: Vec<IssuePattern>,
    /// Layer behavior patterns
    pub layer_patterns: Vec<IssuePattern>,
}

/// Pattern for identifying specific issues.
#[derive(Debug, Clone)]
pub struct IssuePattern {
    /// Pattern name
    pub name: String,
    /// Pattern description
    pub description: String,
    /// Conditions that must be met
    pub conditions: Vec<PatternCondition>,
    /// Associated issue category
    pub issue_category: IssueCategory,
    /// Confidence weight for this pattern
    pub confidence_weight: f64,
    /// Recommended solutions
    pub solutions: Vec<String>,
}

/// Condition for pattern matching.
#[derive(Debug, Clone)]
pub struct PatternCondition {
    /// Metric name
    pub metric: String,
    /// Comparison operator
    pub operator: ComparisonOperator,
    /// Threshold value
    pub threshold: f64,
    /// Number of consecutive occurrences required
    pub consecutive_count: usize,
}

/// Comparison operators for pattern conditions.
#[derive(Debug, Clone)]
pub enum ComparisonOperator {
    /// Greater than
    GreaterThan,
    /// Less than
    LessThan,
    /// Equal to (within tolerance)
    EqualTo,
    /// Increasing trend
    Increasing,
    /// Decreasing trend
    Decreasing,
    /// Oscillating pattern
    Oscillating,
}

/// Session statistics for debugging analysis.
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
pub struct SessionStatistics {
    /// Total issues identified
    pub total_issues: usize,
    /// Issues by category
    pub issues_by_category: HashMap<IssueCategory, usize>,
    /// Total recommendations generated
    pub total_recommendations: usize,
    /// Average confidence of recommendations
    pub avg_recommendation_confidence: f64,
    /// Analysis time taken
    pub analysis_duration: chrono::Duration,
}

impl Default for AutoDebugConfig {
    fn default() -> Self {
        Self {
            max_history_size: 1000,
            min_samples_for_analysis: 10,
            recommendation_confidence_threshold: 0.7,
            enable_advanced_patterns: true,
            enable_hyperparameter_suggestions: true,
            enable_architectural_recommendations: true,
        }
    }
}

impl AutoDebugger {
    /// Create a new auto-debugger.
    pub fn new() -> Self {
        Self {
            config: AutoDebugConfig::default(),
            performance_history: VecDeque::new(),
            layer_history: HashMap::new(),
            dynamics_history: VecDeque::new(),
            issue_patterns: IssuePatternDatabase::new(),
            session_state: DebuggingSession::new(),
        }
    }

    /// Create auto-debugger with custom configuration.
    pub fn with_config(config: AutoDebugConfig) -> Self {
        Self {
            config,
            performance_history: VecDeque::new(),
            layer_history: HashMap::new(),
            dynamics_history: VecDeque::new(),
            issue_patterns: IssuePatternDatabase::new(),
            session_state: DebuggingSession::new(),
        }
    }

    /// Record performance metrics for analysis.
    pub fn record_performance_metrics(&mut self, metrics: ModelPerformanceMetrics) {
        self.performance_history.push_back(metrics);

        while self.performance_history.len() > self.config.max_history_size {
            self.performance_history.pop_front();
        }
    }

    /// Record layer statistics for analysis.
    pub fn record_layer_stats(&mut self, stats: LayerActivationStats) {
        let layer_name = stats.layer_name.clone();

        let layer_history = self.layer_history.entry(layer_name).or_default();
        layer_history.push_back(stats);

        while layer_history.len() > self.config.max_history_size {
            layer_history.pop_front();
        }
    }

    /// Record training dynamics for analysis.
    pub fn record_training_dynamics(&mut self, dynamics: TrainingDynamics) {
        self.dynamics_history.push_back(dynamics);

        while self.dynamics_history.len() > self.config.max_history_size {
            self.dynamics_history.pop_front();
        }
    }

    /// Perform comprehensive auto-debugging analysis.
    pub fn perform_analysis(&mut self) -> Result<DebuggingReport> {
        let analysis_start = chrono::Utc::now();

        if self.performance_history.len() < self.config.min_samples_for_analysis {
            return Err(anyhow::anyhow!("Insufficient data for analysis"));
        }

        // Clear previous session state
        self.session_state = DebuggingSession::new();

        // Analyze different aspects
        self.analyze_learning_rate_issues()?;
        self.analyze_convergence_issues()?;
        self.analyze_gradient_flow_issues()?;
        self.analyze_layer_health_issues()?;
        self.analyze_memory_issues()?;
        self.analyze_overfitting_underfitting()?;

        // Generate recommendations based on identified issues
        self.generate_recommendations()?;

        // Update session statistics
        self.session_state.session_stats.analysis_duration = chrono::Utc::now() - analysis_start;
        self.update_session_statistics();

        Ok(DebuggingReport {
            session_info: self.session_state.clone(),
            identified_issues: self.session_state.identified_issues.clone(),
            recommendations: self.session_state.recommendations.clone(),
            summary: self.generate_analysis_summary(),
        })
    }

    /// Analyze learning rate related issues.
    fn analyze_learning_rate_issues(&mut self) -> Result<()> {
        let recent_metrics: Vec<_> = self.performance_history.iter().rev().take(20).collect();
        if recent_metrics.len() < 10 {
            return Ok(());
        }

        let mut issues_to_add = Vec::new();

        // Check for loss explosion
        let recent_losses: Vec<f64> = recent_metrics.iter().map(|m| m.loss).collect();
        if let Some(max_loss) = recent_losses
            .iter()
            .max_by(|a, b| a.partial_cmp(b).unwrap_or(std::cmp::Ordering::Equal))
        {
            if let Some(min_loss) = recent_losses
                .iter()
                .min_by(|a, b| a.partial_cmp(b).unwrap_or(std::cmp::Ordering::Equal))
            {
                if max_loss / min_loss > 10.0 {
                    issues_to_add.push(IdentifiedIssue {
                        category: IssueCategory::LearningRate,
                        description: "Learning rate too high - loss explosion detected".to_string(),
                        severity: IssueSeverity::Critical,
                        confidence: 0.9,
                        evidence: vec![
                            format!("Loss ratio: {:.2}", max_loss / min_loss),
                            "Rapid loss increase observed".to_string(),
                        ],
                        potential_causes: vec![
                            "Learning rate set too high".to_string(),
                            "Gradient clipping disabled".to_string(),
                            "Numerical instability".to_string(),
                        ],
                        identified_at: chrono::Utc::now(),
                    });
                }
            }
        }

        // Check for learning stagnation
        let loss_variance = self.calculate_variance(&recent_losses);
        let recent_metrics_len = recent_metrics.len();
        if loss_variance < 1e-6 && recent_metrics_len >= 15 {
            issues_to_add.push(IdentifiedIssue {
                category: IssueCategory::LearningRate,
                description: "Learning rate too low - training stagnation".to_string(),
                severity: IssueSeverity::Major,
                confidence: 0.8,
                evidence: vec![
                    format!("Loss variance: {:.2e}", loss_variance),
                    "No learning progress in recent steps".to_string(),
                ],
                potential_causes: vec![
                    "Learning rate set too low".to_string(),
                    "Learning rate decay too aggressive".to_string(),
                    "Model has converged".to_string(),
                ],
                identified_at: chrono::Utc::now(),
            });
        }

        // Add all collected issues
        for issue in issues_to_add {
            self.add_issue(issue);
        }

        Ok(())
    }

    /// Analyze convergence related issues.
    fn analyze_convergence_issues(&mut self) -> Result<()> {
        if let Some(latest_dynamics) = self.dynamics_history.back() {
            match latest_dynamics.convergence_status {
                ConvergenceStatus::Diverging => {
                    self.add_issue(IdentifiedIssue {
                        category: IssueCategory::Convergence,
                        description: "Training is diverging".to_string(),
                        severity: IssueSeverity::Critical,
                        confidence: 0.95,
                        evidence: vec!["Convergence status: Diverging".to_string()],
                        potential_causes: vec![
                            "Learning rate too high".to_string(),
                            "Gradient explosion".to_string(),
                            "Numerical instability".to_string(),
                        ],
                        identified_at: chrono::Utc::now(),
                    });
                },
                ConvergenceStatus::Plateau => {
                    if let Some(plateau_info) = &latest_dynamics.plateau_detection {
                        if plateau_info.duration_steps > 100 {
                            self.add_issue(IdentifiedIssue {
                                category: IssueCategory::Convergence,
                                description: "Training has plateaued".to_string(),
                                severity: IssueSeverity::Moderate,
                                confidence: 0.8,
                                evidence: vec![
                                    format!(
                                        "Plateau duration: {} steps",
                                        plateau_info.duration_steps
                                    ),
                                    format!("Plateau value: {:.4}", plateau_info.plateau_value),
                                ],
                                potential_causes: vec![
                                    "Learning rate too low".to_string(),
                                    "Model capacity insufficient".to_string(),
                                    "Local minimum reached".to_string(),
                                ],
                                identified_at: chrono::Utc::now(),
                            });
                        }
                    }
                },
                ConvergenceStatus::Oscillating => {
                    self.add_issue(IdentifiedIssue {
                        category: IssueCategory::NumericalStability,
                        description: "Training is oscillating".to_string(),
                        severity: IssueSeverity::Moderate,
                        confidence: 0.7,
                        evidence: vec!["Convergence status: Oscillating".to_string()],
                        potential_causes: vec![
                            "Learning rate too high".to_string(),
                            "Batch size too small".to_string(),
                            "Momentum settings suboptimal".to_string(),
                        ],
                        identified_at: chrono::Utc::now(),
                    });
                },
                _ => {},
            }
        }

        Ok(())
    }

    /// Analyze gradient flow issues.
    fn analyze_gradient_flow_issues(&mut self) -> Result<()> {
        let mut issues_to_add = Vec::new();

        // Check layer statistics for gradient flow problems
        for (layer_name, layer_history) in &self.layer_history {
            if let Some(latest_stats) = layer_history.back() {
                // Check for dead neurons
                if latest_stats.dead_neurons_ratio > 0.5 {
                    issues_to_add.push(IdentifiedIssue {
                        category: IssueCategory::GradientFlow,
                        description: format!("High dead neuron ratio in layer {}", layer_name),
                        severity: IssueSeverity::Major,
                        confidence: 0.85,
                        evidence: vec![
                            format!(
                                "Dead neurons: {:.1}%",
                                latest_stats.dead_neurons_ratio * 100.0
                            ),
                            format!("Layer: {}", layer_name),
                        ],
                        potential_causes: vec![
                            "Dying ReLU problem".to_string(),
                            "Poor weight initialization".to_string(),
                            "Learning rate too high".to_string(),
                        ],
                        identified_at: chrono::Utc::now(),
                    });
                }

                // Check for activation saturation
                if latest_stats.saturated_neurons_ratio > 0.3 {
                    issues_to_add.push(IdentifiedIssue {
                        category: IssueCategory::GradientFlow,
                        description: format!("High activation saturation in layer {}", layer_name),
                        severity: IssueSeverity::Moderate,
                        confidence: 0.8,
                        evidence: vec![
                            format!(
                                "Saturated neurons: {:.1}%",
                                latest_stats.saturated_neurons_ratio * 100.0
                            ),
                            format!("Layer: {}", layer_name),
                        ],
                        potential_causes: vec![
                            "Vanishing gradient problem".to_string(),
                            "Poor activation function choice".to_string(),
                            "Input normalization issues".to_string(),
                        ],
                        identified_at: chrono::Utc::now(),
                    });
                }
            }
        }

        // Add all collected issues
        for issue in issues_to_add {
            self.add_issue(issue);
        }

        Ok(())
    }

    /// Analyze layer health issues.
    fn analyze_layer_health_issues(&mut self) -> Result<()> {
        let mut issues_to_add = Vec::new();

        for (layer_name, layer_history) in &self.layer_history {
            if layer_history.len() >= 5 {
                let recent_stats: Vec<_> = layer_history.iter().rev().take(5).collect();

                // Check for activation variance trends
                let variances: Vec<f64> = recent_stats.iter().map(|s| s.std_activation).collect();
                let avg_variance = variances.iter().sum::<f64>() / variances.len() as f64;

                if avg_variance < 0.01 {
                    issues_to_add.push(IdentifiedIssue {
                        category: IssueCategory::Architecture,
                        description: format!("Low activation variance in layer {}", layer_name),
                        severity: IssueSeverity::Minor,
                        confidence: 0.6,
                        evidence: vec![
                            format!("Average variance: {:.4}", avg_variance),
                            format!("Layer: {}", layer_name),
                        ],
                        potential_causes: vec![
                            "Poor weight initialization".to_string(),
                            "Input normalization too aggressive".to_string(),
                            "Activation function saturation".to_string(),
                        ],
                        identified_at: chrono::Utc::now(),
                    });
                }
            }
        }

        // Add all collected issues
        for issue in issues_to_add {
            self.add_issue(issue);
        }

        Ok(())
    }

    /// Analyze memory usage issues.
    fn analyze_memory_issues(&mut self) -> Result<()> {
        if self.performance_history.len() >= 10 {
            let recent_memory: Vec<f64> = self
                .performance_history
                .iter()
                .rev()
                .take(10)
                .map(|m| m.memory_usage_mb)
                .collect();

            // Check for memory leaks
            let memory_trend = self.calculate_trend(&recent_memory);
            if memory_trend > 10.0 {
                // MB per step
                self.add_issue(IdentifiedIssue {
                    category: IssueCategory::Memory,
                    description: "Memory leak detected".to_string(),
                    severity: IssueSeverity::Critical,
                    confidence: 0.9,
                    evidence: vec![
                        format!("Memory growth rate: {:.2} MB/step", memory_trend),
                        "Increasing memory usage trend".to_string(),
                    ],
                    potential_causes: vec![
                        "Gradient accumulation without clearing".to_string(),
                        "Cached tensors not being released".to_string(),
                        "Memory fragmentation".to_string(),
                    ],
                    identified_at: chrono::Utc::now(),
                });
            }

            // Check for excessive memory usage
            if let Some(max_memory) = recent_memory
                .iter()
                .max_by(|a, b| a.partial_cmp(b).unwrap_or(std::cmp::Ordering::Equal))
            {
                if *max_memory > 16384.0 {
                    // 16GB
                    self.add_issue(IdentifiedIssue {
                        category: IssueCategory::Memory,
                        description: "Excessive memory usage detected".to_string(),
                        severity: IssueSeverity::Major,
                        confidence: 0.8,
                        evidence: vec![
                            format!("Peak memory: {:.0} MB", max_memory),
                            "High memory consumption".to_string(),
                        ],
                        potential_causes: vec![
                            "Batch size too large".to_string(),
                            "Model too large for available memory".to_string(),
                            "Inefficient memory allocation".to_string(),
                        ],
                        identified_at: chrono::Utc::now(),
                    });
                }
            }
        }

        Ok(())
    }

    /// Analyze overfitting and underfitting issues.
    fn analyze_overfitting_underfitting(&mut self) -> Result<()> {
        let mut issues_to_add = Vec::new();

        if let Some(latest_dynamics) = self.dynamics_history.back() {
            // Check for overfitting indicators
            for indicator in &latest_dynamics.overfitting_indicators {
                if let super::types::OverfittingIndicator::TrainValidationGap { gap } = indicator {
                    if *gap > 0.1 {
                        issues_to_add.push(IdentifiedIssue {
                            category: IssueCategory::Overfitting,
                            description: "Large training-validation gap detected".to_string(),
                            severity: IssueSeverity::Major,
                            confidence: 0.85,
                            evidence: vec![
                                format!("Train-validation gap: {:.3}", gap),
                                "Overfitting indicator present".to_string(),
                            ],
                            potential_causes: vec![
                                "Model complexity too high".to_string(),
                                "Insufficient regularization".to_string(),
                                "Training set too small".to_string(),
                            ],
                            identified_at: chrono::Utc::now(),
                        });
                    }
                }
            }

            // Check for underfitting indicators
            for indicator in &latest_dynamics.underfitting_indicators {
                match indicator {
                    super::types::UnderfittingIndicator::HighTrainingLoss { loss, threshold } => {
                        issues_to_add.push(IdentifiedIssue {
                            category: IssueCategory::Underfitting,
                            description: "High training loss indicates underfitting".to_string(),
                            severity: IssueSeverity::Moderate,
                            confidence: 0.7,
                            evidence: vec![
                                format!("Training loss: {:.3}", loss),
                                format!("Threshold: {:.3}", threshold),
                            ],
                            potential_causes: vec![
                                "Model capacity too low".to_string(),
                                "Learning rate too low".to_string(),
                                "Insufficient training time".to_string(),
                            ],
                            identified_at: chrono::Utc::now(),
                        });
                    },
                    super::types::UnderfittingIndicator::SlowConvergence {
                        steps_taken,
                        expected,
                    } => {
                        issues_to_add.push(IdentifiedIssue {
                            category: IssueCategory::Underfitting,
                            description: "Slow convergence detected".to_string(),
                            severity: IssueSeverity::Minor,
                            confidence: 0.6,
                            evidence: vec![
                                format!("Steps taken: {}", steps_taken),
                                format!("Expected: {}", expected),
                            ],
                            potential_causes: vec![
                                "Learning rate too conservative".to_string(),
                                "Optimizer choice suboptimal".to_string(),
                                "Poor initialization".to_string(),
                            ],
                            identified_at: chrono::Utc::now(),
                        });
                    },
                    _ => {},
                }
            }
        }

        // Add all collected issues
        for issue in issues_to_add {
            self.add_issue(issue);
        }

        Ok(())
    }

    /// Generate recommendations based on identified issues.
    fn generate_recommendations(&mut self) -> Result<()> {
        for issue in &self.session_state.identified_issues {
            let recommendations = self.generate_recommendations_for_issue(issue);
            self.session_state.recommendations.extend(recommendations);
        }

        // Sort recommendations by priority and confidence
        self.session_state.recommendations.sort_by(|a, b| {
            b.priority
                .partial_cmp(&a.priority)
                .unwrap_or(std::cmp::Ordering::Equal)
                .then(b.confidence.partial_cmp(&a.confidence).unwrap_or(std::cmp::Ordering::Equal))
        });

        Ok(())
    }

    /// Generate specific recommendations for an issue.
    fn generate_recommendations_for_issue(
        &self,
        issue: &IdentifiedIssue,
    ) -> Vec<DebuggingRecommendation> {
        match issue.category {
            IssueCategory::LearningRate => {
                if issue.description.contains("too high") {
                    vec![DebuggingRecommendation {
                        category: RecommendationCategory::HyperparameterTuning,
                        title: "Reduce Learning Rate".to_string(),
                        description: "Lower the learning rate to stabilize training".to_string(),
                        actions: vec![
                            "Reduce learning rate by factor of 2-10".to_string(),
                            "Enable gradient clipping".to_string(),
                            "Consider learning rate scheduling".to_string(),
                        ],
                        expected_impact: "Stabilized training with reduced loss oscillations"
                            .to_string(),
                        confidence: 0.9,
                        priority: AutoDebugRecommendationPriority::High,
                        hyperparameter_suggestions: vec![HyperparameterSuggestion {
                            parameter_name: "learning_rate".to_string(),
                            current_value: None,
                            suggested_value: 0.0001,
                            reasoning: "Reduce to prevent loss explosion".to_string(),
                            expected_effect: "More stable training".to_string(),
                        }],
                    }]
                } else if issue.description.contains("too low") {
                    vec![DebuggingRecommendation {
                        category: RecommendationCategory::HyperparameterTuning,
                        title: "Increase Learning Rate".to_string(),
                        description: "Increase learning rate to improve convergence speed"
                            .to_string(),
                        actions: vec![
                            "Increase learning rate by factor of 2-5".to_string(),
                            "Use learning rate warmup".to_string(),
                            "Consider adaptive learning rate methods".to_string(),
                        ],
                        expected_impact: "Faster convergence and better final performance"
                            .to_string(),
                        confidence: 0.8,
                        priority: AutoDebugRecommendationPriority::Medium,
                        hyperparameter_suggestions: vec![HyperparameterSuggestion {
                            parameter_name: "learning_rate".to_string(),
                            current_value: None,
                            suggested_value: 0.001,
                            reasoning: "Increase to improve learning speed".to_string(),
                            expected_effect: "Faster convergence".to_string(),
                        }],
                    }]
                } else {
                    Vec::new()
                }
            },
            IssueCategory::Memory => {
                vec![DebuggingRecommendation {
                    category: RecommendationCategory::ResourceOptimization,
                    title: "Optimize Memory Usage".to_string(),
                    description: "Implement memory optimization strategies".to_string(),
                    actions: vec![
                        "Reduce batch size".to_string(),
                        "Enable gradient checkpointing".to_string(),
                        "Clear cached tensors regularly".to_string(),
                        "Use mixed precision training".to_string(),
                    ],
                    expected_impact: "Reduced memory consumption and stable training".to_string(),
                    confidence: 0.85,
                    priority: AutoDebugRecommendationPriority::High,
                    hyperparameter_suggestions: vec![HyperparameterSuggestion {
                        parameter_name: "batch_size".to_string(),
                        current_value: None,
                        suggested_value: 16.0,
                        reasoning: "Reduce to lower memory usage".to_string(),
                        expected_effect: "Lower memory consumption".to_string(),
                    }],
                }]
            },
            IssueCategory::Overfitting => {
                vec![DebuggingRecommendation {
                    category: RecommendationCategory::TrainingStrategy,
                    title: "Address Overfitting".to_string(),
                    description: "Implement regularization strategies to reduce overfitting"
                        .to_string(),
                    actions: vec![
                        "Add dropout layers".to_string(),
                        "Increase weight decay".to_string(),
                        "Use data augmentation".to_string(),
                        "Reduce model complexity".to_string(),
                        "Implement early stopping".to_string(),
                    ],
                    expected_impact: "Better generalization and validation performance".to_string(),
                    confidence: 0.8,
                    priority: AutoDebugRecommendationPriority::Medium,
                    hyperparameter_suggestions: vec![HyperparameterSuggestion {
                        parameter_name: "dropout_rate".to_string(),
                        current_value: None,
                        suggested_value: 0.1,
                        reasoning: "Add regularization to reduce overfitting".to_string(),
                        expected_effect: "Better generalization".to_string(),
                    }],
                }]
            },
            IssueCategory::GradientFlow => {
                vec![DebuggingRecommendation {
                    category: RecommendationCategory::ArchitecturalModification,
                    title: "Improve Gradient Flow".to_string(),
                    description: "Address gradient flow issues in the network".to_string(),
                    actions: vec![
                        "Use different activation functions (e.g., Leaky ReLU, Swish)".to_string(),
                        "Add batch normalization".to_string(),
                        "Implement residual connections".to_string(),
                        "Adjust weight initialization".to_string(),
                    ],
                    expected_impact: "Better gradient flow and training stability".to_string(),
                    confidence: 0.75,
                    priority: AutoDebugRecommendationPriority::Medium,
                    hyperparameter_suggestions: Vec::new(),
                }]
            },
            _ => Vec::new(),
        }
    }

    /// Add an issue to the current session.
    fn add_issue(&mut self, issue: IdentifiedIssue) {
        self.session_state.identified_issues.push(issue);
    }

    /// Calculate variance of a sequence of values.
    fn calculate_variance(&self, values: &[f64]) -> f64 {
        if values.len() < 2 {
            return 0.0;
        }

        let mean = values.iter().sum::<f64>() / values.len() as f64;
        let variance =
            values.iter().map(|x| (x - mean).powi(2)).sum::<f64>() / (values.len() - 1) as f64;

        variance
    }

    /// Calculate trend (slope) of a sequence of values.
    fn calculate_trend(&self, values: &[f64]) -> f64 {
        if values.len() < 2 {
            return 0.0;
        }

        let n = values.len() as f64;
        let x_mean = (n - 1.0) / 2.0;
        let y_mean = values.iter().sum::<f64>() / n;

        let numerator: f64 = values
            .iter()
            .enumerate()
            .map(|(i, &y)| (i as f64 - x_mean) * (y - y_mean))
            .sum();

        let denominator: f64 = (0..values.len()).map(|i| (i as f64 - x_mean).powi(2)).sum();

        if denominator == 0.0 {
            0.0
        } else {
            numerator / denominator
        }
    }

    /// Update session statistics.
    fn update_session_statistics(&mut self) {
        let mut issues_by_category = HashMap::new();
        for issue in &self.session_state.identified_issues {
            *issues_by_category.entry(issue.category.clone()).or_insert(0) += 1;
        }

        let avg_confidence = if self.session_state.recommendations.is_empty() {
            0.0
        } else {
            self.session_state.recommendations.iter().map(|r| r.confidence).sum::<f64>()
                / self.session_state.recommendations.len() as f64
        };

        self.session_state.session_stats = SessionStatistics {
            total_issues: self.session_state.identified_issues.len(),
            issues_by_category,
            total_recommendations: self.session_state.recommendations.len(),
            avg_recommendation_confidence: avg_confidence,
            analysis_duration: self.session_state.session_stats.analysis_duration,
        };
    }

    /// Generate analysis summary.
    fn generate_analysis_summary(&self) -> String {
        let critical_issues = self
            .session_state
            .identified_issues
            .iter()
            .filter(|i| i.severity == IssueSeverity::Critical)
            .count();

        let major_issues = self
            .session_state
            .identified_issues
            .iter()
            .filter(|i| i.severity == IssueSeverity::Major)
            .count();

        let high_priority_recommendations = self
            .session_state
            .recommendations
            .iter()
            .filter(|r| r.priority == AutoDebugRecommendationPriority::High)
            .count();

        format!(
            "Auto-debugging analysis completed. Found {} critical issues, {} major issues. \
            Generated {} recommendations with {} high-priority actions. \
            Average recommendation confidence: {:.2}",
            critical_issues,
            major_issues,
            self.session_state.recommendations.len(),
            high_priority_recommendations,
            self.session_state.session_stats.avg_recommendation_confidence
        )
    }
}

/// Comprehensive debugging report.
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
pub struct DebuggingReport {
    /// Session information
    pub session_info: DebuggingSession,
    /// All identified issues
    pub identified_issues: Vec<IdentifiedIssue>,
    /// Generated recommendations
    pub recommendations: Vec<DebuggingRecommendation>,
    /// Analysis summary
    pub summary: String,
}

impl IssuePatternDatabase {
    /// Create a new pattern database with default patterns.
    pub fn new() -> Self {
        Self {
            learning_rate_patterns: Self::create_learning_rate_patterns(),
            gradient_patterns: Self::create_gradient_patterns(),
            convergence_patterns: Self::create_convergence_patterns(),
            layer_patterns: Self::create_layer_patterns(),
        }
    }

    /// Create default learning rate patterns.
    fn create_learning_rate_patterns() -> Vec<IssuePattern> {
        vec![IssuePattern {
            name: "Loss Explosion".to_string(),
            description: "Rapid increase in loss indicating learning rate too high".to_string(),
            conditions: vec![PatternCondition {
                metric: "loss".to_string(),
                operator: ComparisonOperator::Increasing,
                threshold: 2.0,
                consecutive_count: 3,
            }],
            issue_category: IssueCategory::LearningRate,
            confidence_weight: 0.9,
            solutions: vec![
                "Reduce learning rate by factor of 10".to_string(),
                "Enable gradient clipping".to_string(),
            ],
        }]
    }

    /// Create default gradient patterns.
    fn create_gradient_patterns() -> Vec<IssuePattern> {
        vec![]
    }

    /// Create default convergence patterns.
    fn create_convergence_patterns() -> Vec<IssuePattern> {
        vec![]
    }

    /// Create default layer patterns.
    fn create_layer_patterns() -> Vec<IssuePattern> {
        vec![]
    }
}

impl DebuggingSession {
    /// Create a new debugging session.
    fn new() -> Self {
        Self {
            session_start: chrono::Utc::now(),
            identified_issues: Vec::new(),
            recommendations: Vec::new(),
            session_stats: SessionStatistics {
                total_issues: 0,
                issues_by_category: HashMap::new(),
                total_recommendations: 0,
                avg_recommendation_confidence: 0.0,
                analysis_duration: chrono::Duration::zero(),
            },
        }
    }
}

impl Default for AutoDebugger {
    fn default() -> Self {
        Self::new()
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_auto_debugger_creation() {
        let debugger = AutoDebugger::new();
        assert_eq!(debugger.performance_history.len(), 0);
        assert_eq!(debugger.layer_history.len(), 0);
    }

    #[test]
    fn test_issue_identification() {
        let mut debugger = AutoDebugger::new();

        let issue = IdentifiedIssue {
            category: IssueCategory::LearningRate,
            description: "Test issue".to_string(),
            severity: IssueSeverity::Major,
            confidence: 0.8,
            evidence: vec!["Test evidence".to_string()],
            potential_causes: vec!["Test cause".to_string()],
            identified_at: chrono::Utc::now(),
        };

        debugger.add_issue(issue);
        assert_eq!(debugger.session_state.identified_issues.len(), 1);
    }

    #[test]
    fn test_variance_calculation() {
        let debugger = AutoDebugger::new();
        let values = vec![1.0, 2.0, 3.0, 4.0, 5.0];
        let variance = debugger.calculate_variance(&values);
        assert!(variance > 0.0);
    }

    #[test]
    fn test_trend_calculation() {
        let debugger = AutoDebugger::new();
        let increasing_values = vec![1.0, 2.0, 3.0, 4.0, 5.0];
        let trend = debugger.calculate_trend(&increasing_values);
        assert!(trend > 0.0);
    }
}