Skip to content

Conversation

@DavidIkov
Copy link
Collaborator

No description provided.

Comment on lines +20 to +21
add_subdirectory(blifat)
add_subdirectory(altai)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Что-то здесь не то. У сеток различаются нейроны, однако топология почти дублируется.
Зачем столько копипаста?

Comment on lines +28 to +29
add_executable(mnist_learn_altai_example main.cpp construct_network.cpp train.cpp inference.cpp time_string.cpp)
target_link_libraries(mnist_learn_altai_example PRIVATE KNP::BaseFramework::Core ${Boost_LIBRARIES})
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Вероятно, надо сделать один пример, в котором опциями будет выбираться сеть.
Для построения лучше использовать шаблонные функции.
И передавать им тип нейрона.


enum PopIndexes
{
L,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Комментарий, что ещё за "L" (а лучше поменять название)?

Comment on lines +83 to +95
l_neuron.d_h_ = -0.1765261f * 1000;

// stability_resource_change_ratio
l_neuron.stability_change_parameter_ = 0.0497573 / 1000;

// silent synapses
l_neuron.resource_drain_coefficient_ = 27;

// chartime * hebbian_plasticity_chartime_ratio
// l_neuron.dopamine_plasticity_time_ = 3 * 2.72f;

// threshold_excess_weight_dependent
l_neuron.synapse_sum_threshold_coefficient_ = 0.217654;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Немного бы рассказать о том, что это за магические числа, откуда взялись, почему такие, - это пример для пользователя.

Comment on lines +97 to +103
struct PopulationRole
{
PopulationData<ResourceNeuron> pd_;
bool for_inference_;
bool output_;
std::string name_;
};
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Не стоит ли отметить это для фреймворка?

Comment on lines +110 to +111


Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

-пустая строка.

#include <vector>

#include "impl/populations/interface.h"

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+пустая строка.

future_messages.erase(out_iter);
}
}

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+пустая строка.

impl::calculate_projection_multithreaded_interface(
projection, message_in_data, future_messages, step_n, part_start, part_size, mutex);
}

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

-пустая строка.

Comment on lines +143 to 158
auto call_calculate_post_input_state = [](T &pop_ref,
knp::core::messaging::SpikeMessage &message_ref,
size_t start, size_t end, std::mutex &mutex_ref)
{
uint64_t part_end = std::min<uint64_t>(start + end, pop_ref.size());
knp::core::messaging::SpikeMessage buffer_message;
knp::backends::cpu::populations::calculate_post_impact_population_state(
pop_ref, buffer_message, start, part_end);
const std::lock_guard lock(mutex_ref);
message_ref.neuron_indexes_.reserve(
message_ref.neuron_indexes_.size() + buffer_message.neuron_indexes_.size());
message_ref.neuron_indexes_.insert(
message_ref.neuron_indexes_.end(), buffer_message.neuron_indexes_.begin(),
buffer_message.neuron_indexes_.end());
};
calc_pool_->post(
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Тут явно что-то поехало.

Copy link
Contributor

@artiomn artiomn left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Надо разбить на три.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Refactor implementation of working with datasets. Refactor cpu backend. AltAI LIF STDP

2 participants