Recreate network topology
ChiaraBartolozzi opened this issue · comments
Put numbers on the connections and understand which type of connections are used between different types of neurons
| Neuron Population | Number of neurons | Description |
|---|---|---|
| Mossy Fibres (MF) | 100 neurons | send vestibular signal from head position to VN, GoC and GrC |
| Vestibular Nuclei (VN) | 200 neurons | open loop controller that sets the activity of agonist and antagonist eye actuators |
| Golgi Cells (GoC) | 100 neurons | controls the input activity to PC via GrC |
| Granular Cells (GrC - Parallel Fibres (PF)) | 2000 neurons | receives head position and sends to PC via PF |
| Purkinje Cells (PC) | 200 neurons | integrates activity from head position (coming from MF-GoC-GrC-PF) with error signal (head and eye position) |
| Climbing Fibres (CF or IO) | 200 neurons | sends error signal to PF |
Connectivity:
| Input | Output | Type of connection | Synapse type | Receptor type | number of synapses |
|---|---|---|---|---|---|
| MF (100) | VN (200) | AllToAll | stdp_cos, change of weight between MF-VN is triggered @ spike from MF, depending on temporal distance between PC and MF | AMPA | 20000 |
| MF(100) | GoC(100) | OneToOne | static | 100 | |
| MF (100) | GrC (subpopulation 20) | AllToAll | static | 2000 | |
| GoC (100) | GrC (2 x subpop <40) | All ToAll | static | 8000 | |
| PC (200) | VN (200) | OneToOne | static | GABA | 200 |
| PC (200) | VN (200) | OneToOne | static, w=0 | TEACHING_SIGNAL | 200 |
| GrC (2000) | PC(200) | AllToAll | stdp_sin, change of weight between PF(GrC)-PC is triggered @ spike from PF, depending on temporal distance between PF and CF | AMPA | 400000 |
| CF (200) | PC(200) | OneToOne | static, w=0 | COMPLEX_SPIKE | 200 |
One interesting thing I saw here was static connections with w=0. Does this mean weight = 0?
right! these are the teaching signals, they do not provoke any effect on the postsynaptic neuron, but they trigger the learning
Key numbers to follow
Each mossy fiber (MFs) connected to 20 GrCs to 1 GoCs and to 200 VN cells (excitatory connections).
Each Golgi cell (GoCs) connected to 80 GrCs (inhibitory connections).
Each Granular cells (GrCs) connected to 200 PCs (excitatory connections).
Each Climbing fiber (CFs) connected to one PC each.
Each Purkinje cells (PCs): connected to one VN cell (inhibitory connections).
And connectivity pattern
Mossy 0 to Granule 0 -19, connection delay between 1-10 ms random, weight range [0.00048 0.00072]]
Mossy 1 to Granule 20-39, connection delay between 1-10 ms random, weight range [0.00048 0.00072]]
Mossy 2 to Granule 40-59, connection delay between 1-10 ms random, weight range [0.00048 0.00072]]
Mossy 0 to Golgi 0, connection delay between 1-10 ms random, weight range [0.00016 0.00024]
Mossy 1 to Golgi 1, connection delay between 1-10 ms random, weight range [0.00016 0.00024]
Mossy 1 to Golgi 2, connection delay between 1-10 ms random, weight range [0.00016 0.00024]
Golgi 0 to Granule 20-60
Golgi 1 to Granule 0-20 & Granule 40-80
Golgi 2 to Granule 0-40 & Granule 60-100
Golgi 3 to Granule 20-60 & Granule 80-120
…
…
Golgi 98 to Granule 1920-1960 & 1980-2000
Golgi 99 to Granule 1940-1980
Implemented in cerebellum.py