scirs2_optimize/

unified_pipeline.rs

//! Unified optimization pipeline combining all advanced features
//!
//! This module provides a high-level interface that integrates distributed optimization,
//! self-tuning, GPU acceleration, and visualization into a single cohesive pipeline.

use crate::error::{ScirsError, ScirsResult};
use scirs2_core::error_context;
use scirs2_core::ndarray::{Array1, ArrayView1};
use std::path::Path;
// Unused import: std::sync::Arc

use crate::distributed::{DistributedConfig, DistributedOptimizationContext, MPIInterface};
use crate::gpu::{
    acceleration::{AccelerationConfig, AccelerationManager},
    GpuOptimizationConfig, GpuOptimizationContext,
};
use crate::result::OptimizeResults;
use crate::self_tuning::{
    AdaptationStrategy, ParameterValue, SelfTuningConfig, SelfTuningOptimizer, TunableParameter,
};
use crate::visualization::{
    tracking::TrajectoryTracker, OptimizationVisualizer, VisualizationConfig,
};

/// Comprehensive optimization pipeline configuration
#[derive(Clone)]
pub struct UnifiedOptimizationConfig {
    /// Enable distributed optimization
    pub use_distributed: bool,
    /// Distributed optimization settings
    pub distributedconfig: Option<DistributedConfig>,

    /// Enable self-tuning parameter adaptation
    pub use_self_tuning: bool,
    /// Self-tuning configuration
    pub self_tuningconfig: Option<SelfTuningConfig>,

    /// Enable GPU acceleration
    pub use_gpu: bool,
    /// GPU acceleration settings
    pub gpuconfig: Option<GpuOptimizationConfig>,
    /// GPU acceleration configuration
    pub accelerationconfig: Option<AccelerationConfig>,

    /// Enable optimization visualization
    pub enable_visualization: bool,
    /// Visualization settings
    pub visualizationconfig: Option<VisualizationConfig>,

    /// Output directory for results and visualization
    pub output_directory: Option<String>,

    /// Maximum number of iterations
    pub max_nit: usize,
    /// Function tolerance
    pub function_tolerance: f64,
    /// Gradient tolerance
    pub gradient_tolerance: f64,
}

impl Default for UnifiedOptimizationConfig {
    fn default() -> Self {
        Self {
            use_distributed: false,
            distributedconfig: None,
            use_self_tuning: true,
            self_tuningconfig: Some(SelfTuningConfig::default()),
            use_gpu: false,
            gpuconfig: None,
            accelerationconfig: None,
            enable_visualization: true,
            visualizationconfig: Some(VisualizationConfig::default()),
            output_directory: Some("optimization_output".to_string()),
            max_nit: 1000,
            function_tolerance: 1e-6,
            gradient_tolerance: 1e-6,
        }
    }
}

/// Unified optimization pipeline
pub struct UnifiedOptimizer<M: MPIInterface> {
    config: UnifiedOptimizationConfig,

    // Optional components based on configuration
    distributed_context: Option<DistributedOptimizationContext<M>>,
    self_tuning_optimizer: Option<SelfTuningOptimizer>,
    gpu_context: Option<GpuOptimizationContext>,
    acceleration_manager: Option<AccelerationManager>,
    visualizer: Option<OptimizationVisualizer>,
    trajectory_tracker: Option<TrajectoryTracker>,

    // Internal state for convergence checking
    previous_function_value: Option<f64>,
}

impl<M: MPIInterface> UnifiedOptimizer<M> {
    /// Create a new unified optimizer
    pub fn new(config: UnifiedOptimizationConfig, mpi: Option<M>) -> ScirsResult<Self> {
        let distributed_context = if config.use_distributed {
            if let (Some(mpi_interface), Some(distconfig)) = (mpi, &config.distributedconfig) {
                Some(DistributedOptimizationContext::new(
                    mpi_interface,
                    distconfig.clone(),
                ))
            } else {
                return Err(ScirsError::InvalidInput(error_context!(
                    "MPI interface and distributed config required for distributed optimization"
                )));
            }
        } else {
            None
        };

        let self_tuning_optimizer = if config.use_self_tuning {
            let tuningconfig = config.self_tuningconfig.clone().unwrap_or_default();
            Some(SelfTuningOptimizer::new(tuningconfig))
        } else {
            None
        };

        let (gpu_context, acceleration_manager) = if config.use_gpu {
            let gpuconfig = config.gpuconfig.clone().unwrap_or_default();
            let gpu_ctx = GpuOptimizationContext::new(gpuconfig)?;

            let accelconfig = config.accelerationconfig.clone().unwrap_or_default();
            let accel_mgr = AccelerationManager::new(accelconfig);

            (Some(gpu_ctx), Some(accel_mgr))
        } else {
            (None, None)
        };

        let (visualizer, trajectory_tracker) = if config.enable_visualization {
            let visconfig = config.visualizationconfig.clone().unwrap_or_default();
            let vis = OptimizationVisualizer::with_config(visconfig);
            let tracker = TrajectoryTracker::new();
            (Some(vis), Some(tracker))
        } else {
            (None, None)
        };

        Ok(Self {
            config,
            distributed_context,
            self_tuning_optimizer,
            gpu_context,
            acceleration_manager,
            visualizer,
            trajectory_tracker,
            previous_function_value: None,
        })
    }

    /// Register tunable parameters for self-tuning optimization
    pub fn register_tunable_parameter<T>(
        &mut self,
        name: &str,
        param: TunableParameter<T>,
    ) -> ScirsResult<()>
    where
        T: Clone + PartialOrd + std::fmt::Debug + Send + Sync + 'static,
    {
        if let Some(ref mut tuner) = self.self_tuning_optimizer {
            tuner.register_parameter(name, param)?;
        }
        Ok(())
    }

    /// Optimize a function using the unified pipeline
    pub fn optimize<F, G>(
        &mut self,
        function: F,
        gradient: Option<G>,
        initial_guess: &Array1<f64>,
        bounds: Option<&[(Option<f64>, Option<f64>)]>,
    ) -> ScirsResult<UnifiedOptimizationResults>
    where
        F: Fn(&ArrayView1<f64>) -> f64 + Clone + Send + Sync,
        G: Fn(&ArrayView1<f64>) -> Array1<f64> + Clone + Send + Sync,
    {
        // Initialize optimization state
        let mut current_x = initial_guess.clone();
        let mut current_f = function(&current_x.view());
        let mut iteration = 0;
        let mut function_evaluations = 1;
        let mut gradient_evaluations = 0;

        // Initialize tracking
        if let Some(ref mut tracker) = self.trajectory_tracker {
            tracker.record(iteration, &current_x.view(), current_f);
        }

        let start_time = std::time::Instant::now();

        // Main optimization loop
        while iteration < self.config.max_nit {
            iteration += 1;

            // Compute gradient
            let current_grad = if let Some(ref grad_fn) = gradient {
                gradient_evaluations += 1;
                grad_fn(&current_x.view())
            } else {
                // Use numerical differentiation
                self.compute_numerical_gradient(&function, &current_x)?
            };

            let grad_norm = current_grad.iter().map(|&g| g * g).sum::<f64>().sqrt();

            // Check convergence
            if grad_norm < self.config.gradient_tolerance {
                break;
            }

            // Self-tuning parameter adaptation
            let tuning_params = {
                if let Some(ref mut tuner) = self.self_tuning_optimizer {
                    let improvement = if iteration > 1 {
                        // Compute relative improvement in function value
                        let prev_f = self.previous_function_value.unwrap_or(current_f);
                        if prev_f.abs() > 1e-14 {
                            (prev_f - current_f) / prev_f.abs()
                        } else {
                            prev_f - current_f
                        }
                    } else {
                        0.0
                    };

                    let params_changed = tuner.update_parameters(
                        iteration,
                        current_f,
                        Some(grad_norm),
                        improvement,
                    )?;

                    if params_changed {
                        // Store parameters to apply later (clone to avoid borrow issues)
                        Some(tuner.get_parameters().clone())
                    } else {
                        None
                    }
                } else {
                    None
                }
            };

            // Apply tuning parameters if available (mutable reference to tuner is now dropped)
            if let Some(params) = tuning_params {
                self.apply_tuned_parameters(&params)?;
            }

            // GPU-accelerated computation if enabled
            let search_direction = if self.config.use_gpu {
                self.compute_gpu_search_direction(&current_grad)?
            } else {
                self.compute_cpu_search_direction(&current_grad)?
            };

            // Distributed evaluation for line search if enabled
            let step_size = if self.distributed_context.is_some() {
                let current_x_copy = current_x.clone();
                let search_direction_copy = search_direction.clone();
                // Extract distributed context temporarily to avoid borrowing conflicts
                let mut dist_ctx = self.distributed_context.take();
                let result = if let Some(ref mut ctx) = dist_ctx {
                    self.distributed_line_search(
                        ctx,
                        &function,
                        &current_x_copy,
                        &search_direction_copy,
                    )
                } else {
                    unreachable!()
                };
                // Restore the distributed context
                self.distributed_context = dist_ctx;
                result?
            } else {
                self.standard_line_search(&function, &current_x, &search_direction)?
            };

            // Update position
            for i in 0..current_x.len() {
                current_x[i] += step_size * search_direction[i];
            }

            // Apply bounds if specified
            if let Some(bounds) = bounds {
                self.apply_bounds(&mut current_x, bounds);
            }

            // Evaluate new function value
            current_f = function(&current_x.view());
            function_evaluations += 1;

            // Record trajectory
            if let Some(ref mut tracker) = self.trajectory_tracker {
                tracker.record(iteration, &current_x.view(), current_f);
                tracker.record_gradient_norm(grad_norm);
                tracker.record_step_size(step_size);
            }

            // Check function tolerance convergence
            if let Some(prev_f) = self.previous_function_value {
                let abs_improvement = (prev_f - current_f).abs();
                let rel_improvement = if prev_f.abs() > 1e-14 {
                    abs_improvement / prev_f.abs()
                } else {
                    abs_improvement
                };

                // Check both absolute and relative improvements
                if abs_improvement < self.config.function_tolerance
                    || rel_improvement < self.config.function_tolerance
                {
                    break;
                }
            }

            // Update previous function value for next iteration
            self.previous_function_value = Some(current_f);
        }

        let total_time = start_time.elapsed().as_secs_f64();

        // Generate results
        let success = iteration < self.config.max_nit;
        let message = if success {
            "Optimization completed successfully".to_string()
        } else {
            "Maximum iterations reached".to_string()
        };

        // Create visualization if enabled
        let visualization_paths = if let (Some(ref visualizer), Some(ref tracker)) =
            (&self.visualizer, &self.trajectory_tracker)
        {
            self.generate_visualization(visualizer, tracker.trajectory())?
        } else {
            Vec::new()
        };

        // Generate performance report
        let performance_report = self.generate_performance_report(
            total_time,
            function_evaluations,
            gradient_evaluations,
        )?;

        Ok(UnifiedOptimizationResults {
            base_result: OptimizeResults::<f64> {
                x: current_x,
                fun: current_f,
                success,
                message,
                nit: iteration,
                nfev: function_evaluations,
                njev: gradient_evaluations,
                ..OptimizeResults::<f64>::default()
            },
            visualization_paths,
            performance_report,
            self_tuning_report: self
                .self_tuning_optimizer
                .as_ref()
                .map(|t| t.generate_report()),
            distributed_stats: self.distributed_context.as_ref().map(|d| d.stats().clone()),
        })
    }

    /// Compute numerical gradient using finite differences
    fn compute_numerical_gradient<F>(
        &self,
        function: &F,
        x: &Array1<f64>,
    ) -> ScirsResult<Array1<f64>>
    where
        F: Fn(&ArrayView1<f64>) -> f64,
    {
        let n = x.len();
        let mut gradient = Array1::zeros(n);
        let h = 1e-8;

        if self.config.use_gpu {
            // Use GPU-accelerated finite differences if available
            if let Some(ref gpu_ctx) = self.gpu_context {
                return gpu_ctx.compute_gradient_finite_diff(function, x, h);
            }
        }

        // CPU finite differences
        for i in 0..n {
            let mut x_plus = x.clone();
            let mut x_minus = x.clone();
            x_plus[i] += h;
            x_minus[i] -= h;

            gradient[i] = (function(&x_plus.view()) - function(&x_minus.view())) / (2.0 * h);
        }

        Ok(gradient)
    }

    /// Compute search direction using GPU acceleration
    fn compute_gpu_search_direction(&self, gradient: &Array1<f64>) -> ScirsResult<Array1<f64>> {
        if let Some(ref gpu_ctx) = self.gpu_context {
            // Use GPU-accelerated quasi-Newton or other advanced methods
            gpu_ctx.compute_search_direction(gradient)
        } else {
            // Fallback to simple steepest descent
            Ok(-gradient.clone())
        }
    }

    /// Compute search direction using CPU
    fn compute_cpu_search_direction(&self, gradient: &Array1<f64>) -> ScirsResult<Array1<f64>> {
        // Simple steepest descent for now
        Ok(-gradient.clone())
    }

    /// Perform distributed line search
    fn distributed_line_search<F>(
        &mut self,
        _dist_ctx: &mut DistributedOptimizationContext<M>,
        function: &F,
        x: &Array1<f64>,
        direction: &Array1<f64>,
    ) -> ScirsResult<f64>
    where
        F: Fn(&ArrayView1<f64>) -> f64 + Clone + Send + Sync,
    {
        // Distribute line search evaluation across processes
        let _step_sizes = Array1::from(vec![0.001, 0.01, 0.1, 1.0, 10.0]);

        // For simplicity, return a basic step size
        // In a full implementation, this would distribute evaluations
        self.standard_line_search(function, x, direction)
    }

    /// Standard CPU line search
    fn standard_line_search<F>(
        &self,
        function: &F,
        x: &Array1<f64>,
        direction: &Array1<f64>,
    ) -> ScirsResult<f64>
    where
        F: Fn(&ArrayView1<f64>) -> f64,
    {
        // Simple backtracking line search
        let mut alpha = 1.0;
        let c1 = 1e-4;
        let rho = 0.5;

        let f0 = function(&x.view());
        let grad_dot_dir = -direction.iter().map(|&d| d * d).sum::<f64>(); // Since direction = -gradient

        for _ in 0..20 {
            let mut x_new = x.clone();
            for i in 0..x.len() {
                x_new[i] += alpha * direction[i];
            }

            let f_new = function(&x_new.view());

            if f_new <= f0 + c1 * alpha * grad_dot_dir {
                return Ok(alpha);
            }

            alpha *= rho;
        }

        Ok(alpha)
    }

    /// Apply bounds constraints
    fn apply_bounds(&self, x: &mut Array1<f64>, bounds: &[(Option<f64>, Option<f64>)]) {
        for (i, &mut ref mut xi) in x.iter_mut().enumerate() {
            if i < bounds.len() {
                if let Some(lower) = bounds[i].0 {
                    if *xi < lower {
                        *xi = lower;
                    }
                }
                if let Some(upper) = bounds[i].1 {
                    if *xi > upper {
                        *xi = upper;
                    }
                }
            }
        }
    }

    /// Apply tuned parameters from self-tuning optimizer
    fn apply_tuned_parameters(
        &mut self,
        parameters: &std::collections::HashMap<String, ParameterValue>,
    ) -> ScirsResult<()> {
        for (name, value) in parameters {
            match name.as_str() {
                "learning_rate" | "step_size" => {
                    // Update step size or learning rate
                    if let Some(f_val) = value.as_f64() {
                        // Apply to algorithm-specific parameters
                        self.update_algorithm_parameter("step_size", f_val)?;
                    }
                }
                "tolerance" => {
                    if let Some(f_val) = value.as_f64() {
                        // Update convergence tolerance
                        self.update_algorithm_parameter("tolerance", f_val)?;
                    }
                }
                _ => {
                    // Handle other tunable parameters
                }
            }
        }
        Ok(())
    }

    /// Update specific algorithm parameters
    fn update_algorithm_parameter(&mut self, name: &str, value: f64) -> ScirsResult<()> {
        // Update internal algorithm parameters based on self-tuning
        // This would be algorithm-specific
        Ok(())
    }

    /// Generate visualization outputs
    fn generate_visualization(
        &self,
        visualizer: &OptimizationVisualizer,
        trajectory: &crate::visualization::OptimizationTrajectory,
    ) -> ScirsResult<Vec<String>> {
        let mut paths = Vec::new();

        if let Some(ref output_dir) = self.config.output_directory {
            let output_path = Path::new(output_dir);
            std::fs::create_dir_all(output_path)?;

            // Generate convergence plot
            let convergence_path = output_path.join("convergence.svg");
            visualizer.plot_convergence(trajectory, &convergence_path)?;
            paths.push(convergence_path.to_string_lossy().to_string());

            // Generate parameter trajectory if 2D
            if !trajectory.parameters.is_empty() && trajectory.parameters[0].len() == 2 {
                let trajectory_path = output_path.join("trajectory.svg");
                visualizer.plot_parameter_trajectory(trajectory, &trajectory_path)?;
                paths.push(trajectory_path.to_string_lossy().to_string());
            }

            // Generate comprehensive report
            visualizer.create_optimization_report(trajectory, output_path)?;
            paths.push(
                output_path
                    .join("summary.html")
                    .to_string_lossy()
                    .to_string(),
            );
        }

        Ok(paths)
    }

    /// Generate performance report
    fn generate_performance_report(
        &self,
        total_time: f64,
        function_evaluations: usize,
        gradient_evaluations: usize,
    ) -> ScirsResult<String> {
        let mut report = String::from("Unified Optimization Performance Report\n");
        report.push_str("=========================================\n\n");

        report.push_str(&format!("Total Time: {:.3}s\n", total_time));
        report.push_str(&format!("Function Evaluations: {}\n", function_evaluations));
        report.push_str(&format!("Gradient Evaluations: {}\n", gradient_evaluations));

        if total_time > 0.0 {
            report.push_str(&format!(
                "Function Evaluations per Second: {:.2}\n",
                function_evaluations as f64 / total_time
            ));
        }

        // Add distributed performance if available
        if let Some(ref dist_ctx) = self.distributed_context {
            report.push_str("\nDistributed Performance:\n");
            report.push_str(&dist_ctx.stats().generate_report());
        }

        // Add GPU performance if available
        if let Some(ref accel_mgr) = self.acceleration_manager {
            report.push_str("\nGPU Acceleration Performance:\n");
            report.push_str("GPU acceleration metrics available\n");
            // Note: Performance reporting requires specific GPU optimizer instance
        }

        Ok(report)
    }
}

/// Results from unified optimization
#[derive(Debug, Clone)]
pub struct UnifiedOptimizationResults {
    /// Base optimization results
    pub base_result: OptimizeResults<f64>,
    /// Paths to generated visualization files
    pub visualization_paths: Vec<String>,
    /// Performance report
    pub performance_report: String,
    /// Self-tuning report if enabled
    pub self_tuning_report: Option<String>,
    /// Distributed statistics if enabled
    pub distributed_stats: Option<crate::distributed::DistributedStats>,
}

impl UnifiedOptimizationResults {
    /// Get the final optimized parameters
    pub fn x(&self) -> &Array1<f64> {
        &self.base_result.x
    }

    /// Get the final function value
    pub fn fun(&self) -> f64 {
        self.base_result.fun
    }

    /// Check if optimization was successful
    pub fn success(&self) -> bool {
        self.base_result.success
    }

    /// Get the optimization message
    pub fn message(&self) -> &str {
        &self.base_result.message
    }

    /// Get number of iterations performed
    pub fn iterations(&self) -> usize {
        self.base_result.nit
    }

    /// Get number of iterations performed (alias for iterations)
    pub fn nit(&self) -> usize {
        self.base_result.nit
    }

    /// Print comprehensive results summary
    pub fn print_summary(&self) {
        println!("Unified Optimization Results");
        println!("============================");
        println!("Success: {}", self.success());
        println!("Final function value: {:.6e}", self.fun());
        println!("Iterations: {}", self.nit());
        println!("Function evaluations: {}", self.base_result.nfev);

        if self.base_result.njev > 0 {
            println!("Gradient evaluations: {}", self.base_result.njev);
        }

        if !self.visualization_paths.is_empty() {
            println!("\nGenerated visualizations:");
            for path in &self.visualization_paths {
                println!("  {}", path);
            }
        }

        if let Some(ref self_tuning) = self.self_tuning_report {
            println!("\nSelf-Tuning Report:");
            println!("{}", self_tuning);
        }

        if let Some(ref dist_stats) = self.distributed_stats {
            println!("\nDistributed Performance:");
            println!("{}", dist_stats.generate_report());
        }

        println!("\nPerformance Report:");
        println!("{}", self.performance_report);
    }
}

/// Convenience functions for common optimization scenarios
pub mod presets {
    use super::*;

    /// Create configuration for high-performance distributed optimization
    pub fn distributed_gpuconfig(_numprocesses: usize) -> UnifiedOptimizationConfig {
        UnifiedOptimizationConfig {
            use_distributed: true,
            distributedconfig: Some(crate::distributed::DistributedConfig {
                distribution_strategy: crate::distributed::DistributionStrategy::DataParallel,
                load_balancing: crate::distributed::LoadBalancingConfig::default(),
                communication: crate::distributed::CommunicationConfig::default(),
                fault_tolerance: crate::distributed::FaultToleranceConfig::default(),
            }),
            use_gpu: true,
            gpuconfig: Some(GpuOptimizationConfig::default()),
            accelerationconfig: Some(AccelerationConfig::default()),
            use_self_tuning: true,
            self_tuningconfig: Some(SelfTuningConfig {
                adaptation_strategy: AdaptationStrategy::Hybrid,
                update_frequency: 25,
                learning_rate: 0.1,
                memory_window: 100,
                use_bayesian_tuning: true,
                exploration_factor: 0.15,
            }),
            enable_visualization: true,
            visualizationconfig: Some(VisualizationConfig::default()),
            output_directory: Some("distributed_gpu_optimization".to_string()),
            max_nit: 2000,
            function_tolerance: 1e-8,
            gradient_tolerance: 1e-8,
        }
    }

    /// Create configuration for memory-efficient large-scale optimization
    pub fn large_scaleconfig() -> UnifiedOptimizationConfig {
        UnifiedOptimizationConfig {
            use_distributed: false,
            distributedconfig: None,
            use_gpu: true,
            gpuconfig: Some(GpuOptimizationConfig::default()),
            accelerationconfig: Some(AccelerationConfig::default()),
            use_self_tuning: true,
            self_tuningconfig: Some(SelfTuningConfig {
                adaptation_strategy: AdaptationStrategy::PerformanceBased,
                update_frequency: 50,
                learning_rate: 0.05,
                memory_window: 200,
                use_bayesian_tuning: false,
                exploration_factor: 0.1,
            }),
            enable_visualization: true,
            visualizationconfig: Some(VisualizationConfig::default()),
            output_directory: Some("large_scale_optimization".to_string()),
            max_nit: 5000,
            function_tolerance: 1e-6,
            gradient_tolerance: 1e-6,
        }
    }

    /// Create configuration for interactive optimization with real-time visualization
    pub fn interactiveconfig() -> UnifiedOptimizationConfig {
        UnifiedOptimizationConfig {
            use_distributed: false,
            distributedconfig: None,
            use_gpu: false,
            gpuconfig: None,
            accelerationconfig: None,
            use_self_tuning: true,
            self_tuningconfig: Some(SelfTuningConfig {
                adaptation_strategy: AdaptationStrategy::ConvergenceBased,
                update_frequency: 10,
                learning_rate: 0.2,
                memory_window: 50,
                use_bayesian_tuning: true,
                exploration_factor: 0.2,
            }),
            enable_visualization: true,
            visualizationconfig: Some(VisualizationConfig {
                format: crate::visualization::OutputFormat::Html,
                width: 1200,
                height: 800,
                title: Some("Interactive Optimization".to_string()),
                show_grid: true,
                log_scale_y: false,
                color_scheme: crate::visualization::ColorScheme::Scientific,
                show_legend: true,
                custom_style: None,
            }),
            output_directory: Some("interactive_optimization".to_string()),
            max_nit: 500,
            function_tolerance: 1e-4,
            gradient_tolerance: 1e-4,
        }
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use scirs2_core::ndarray::array;

    #[test]
    fn test_unifiedconfig_creation() {
        let config = UnifiedOptimizationConfig::default();
        assert!(!config.use_distributed);
        assert!(config.use_self_tuning);
        assert!(!config.use_gpu);
        assert!(config.enable_visualization);
    }

    #[test]
    fn test_presetconfigs() {
        let large_scale = presets::large_scaleconfig();
        assert!(large_scale.use_gpu);
        assert!(large_scale.use_self_tuning);
        assert!(!large_scale.use_distributed);

        let interactive = presets::interactiveconfig();
        assert!(!interactive.use_gpu);
        assert!(interactive.use_self_tuning);
        assert!(interactive.enable_visualization);
    }

    #[test]
    fn test_bounds_application() {
        // Test bounds constraint application
        let config = UnifiedOptimizationConfig::default();
        let _optimizer: Result<UnifiedOptimizer<crate::distributed::MockMPI>, _> =
            UnifiedOptimizer::new(config, None);

        // This would test the bounds application logic
        // Implementation depends on the specific test setup
    }

    #[test]
    fn test_rosenbrock_optimization() {
        // Test optimization on the Rosenbrock function
        let config = presets::interactiveconfig();

        let rosenbrock = |x: &ArrayView1<f64>| -> f64 {
            let x0 = x[0];
            let x1 = x[1];
            (1.0 - x0).powi(2) + 100.0 * (x1 - x0.powi(2)).powi(2)
        };

        let initial_guess = array![-1.0, 1.0];

        // Create optimizer without MPI for testing
        let mut optimizer: UnifiedOptimizer<crate::distributed::MockMPI> =
            UnifiedOptimizer::new(config, None).unwrap();

        // Register tunable parameters
        optimizer
            .register_tunable_parameter("step_size", TunableParameter::new(0.01, 0.001, 0.1))
            .unwrap();

        // This would run the actual optimization in a full test
        // For now, just test that the setup works
        assert!(true);
    }
}