8f43927575
examples/ is based on KasperskyLab/knp at 55470ed6 with local changes to the mnist-learn altai/blifat models, cuda-backend CMake and model_loader. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
101 lines
3.6 KiB
C++
101 lines
3.6 KiB
C++
/**
|
|
* @file hyperparameters.h
|
|
* @brief Network hyperparameters for AltAI neuron model.
|
|
* @details These parameters are specifically tuned for the AltAI spiking neural network implementation and include both
|
|
* standard neural parameters and model-specific scaling factors for fixed-point arithmetic optimization.
|
|
* Some parameters are multiplied by 1000 because the AltAI model uses fixed-point arithmetic instead of floating-point
|
|
* operations for computational efficiency. This scaling allows the model to operate effectively with integer-based
|
|
* calculations while maintaining necessary precision for learning and spiking dynamics.
|
|
* @kaspersky_support D. Postnikov
|
|
* @date 28.07.2025
|
|
* @license Apache 2.0
|
|
* @copyright © 2025 AO Kaspersky Lab
|
|
*
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*/
|
|
#pragma once
|
|
|
|
#include <cstdint>
|
|
#include <cstdlib>
|
|
|
|
// cppcheck-suppress missingInclude
|
|
#include "global_config.h"
|
|
|
|
|
|
/**
|
|
* @brief Number of neurons reserved per single digit class in the input population.
|
|
*
|
|
* @details This parameter determines the columnar organization of the input population, with 20 neurons allocated for
|
|
* each of the 10 digit classes (200 total input neurons). Each neuron in a column corresponds to a specific feature
|
|
* of the digit class.
|
|
*/
|
|
// Diagnostics (not part of upstream example): KNP_COL_SIZE=<n> overrides the number of neurons per column.
|
|
inline const size_t neurons_per_column = []
|
|
{
|
|
const char *value = std::getenv("KNP_COL_SIZE");
|
|
return (value && *value) ? static_cast<size_t>(std::atoi(value)) : static_cast<size_t>(20);
|
|
}();
|
|
|
|
|
|
/**
|
|
* @brief the threshold value of neuron potential, after exceeding which a positive spike can be emitted.
|
|
*/
|
|
constexpr uint16_t activation_threshold = 8531;
|
|
/**
|
|
* @brief The constant leakage of the neuron potential.
|
|
*/
|
|
constexpr int16_t potential_leak = 0;
|
|
/**
|
|
* @brief The threshold value of neuron potential, below which a negative spike can be emitted
|
|
*/
|
|
constexpr uint16_t negative_activation_threshold = 0;
|
|
/**
|
|
* @brief A reset value of the neuron potential after one of the thresholds has been exceeded.
|
|
*/
|
|
constexpr uint16_t potential_reset_value = 0;
|
|
/**
|
|
* @brief Time parameter for dopamine plasticity.
|
|
*/
|
|
constexpr uint32_t dopamine_plasticity_time = 10;
|
|
/**
|
|
* @brief Time between spikes in the ISI period.
|
|
*/
|
|
constexpr uint32_t isi_max = 10;
|
|
/**
|
|
* @brief Hebbian plasticity value.
|
|
*/
|
|
constexpr float d_h = -0.1765261f * 1000;
|
|
/**
|
|
* @brief Stability fluctuation value.
|
|
*/
|
|
constexpr float stability_change_parameter = 0.0497573 / 1000;
|
|
/**
|
|
* @brief Number of silent synapses.
|
|
*/
|
|
constexpr uint32_t resource_drain_coefficient = 27;
|
|
/**
|
|
* @brief Synapse sum threshold coefficient.
|
|
*/
|
|
constexpr float synapse_sum_threshold_coefficient = 0.217654;
|
|
/**
|
|
* @brief Dopamine plasticity period for raster to input synapses.
|
|
*/
|
|
constexpr uint32_t raster_to_input_synapse_dopamine_plasticity_period = 10;
|
|
/**
|
|
* @brief Minimum synaptic weight for raster to input projections.
|
|
*/
|
|
constexpr float raster_to_input_synapse_w_min = -0.253122 * 1000;
|
|
/**
|
|
* @brief Maximum synaptic weight for raster to input projections.
|
|
*/
|
|
constexpr float raster_to_input_synapse_w_max = 0.0923957 * 1000;
|