-
Notifications
You must be signed in to change notification settings - Fork 34
Expand file tree
/
Copy pathhistory.txt
More file actions
104 lines (91 loc) · 5.54 KB
/
history.txt
File metadata and controls
104 lines (91 loc) · 5.54 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
=======
History
=======
0.1.0 (2022–05–05)
— — — — — — — — -
* First public release.
0.2.0 (2022–05–09)
— — — — — — — — -
* Major updated release.
* Changes include: edits to API, inclusion/writing of
demonstrations 4 and 5, and a revamped engine core to utilize a configurable
static graph optimization/framework to obtain a 5-7.5 times speedup of
simulated neural dynamics.
0.3.0-alpha (2022–05–21)
— — — — — — — — -
* NGCGraph .compile() further tweaked to use an injection/clamping look-up
system to allow for dynamic changes to occur w/in a static graph compiled
simulated NGC system
* Cable API slightly modified to increase flexibility (demonstrations and tests modified to reflect updated API)
* Demonstration 6 released showcasing how to use ngc-learn to construct/fit a
restricted Boltzmann machine
0.4.0-alpha (2022–05–25)
— — — — — — — — -
* Harmonium written into model museum
* "demonstrations" have been renamed to "walkthroughs"
* tutorial 1 written/provided
* nodes-and-cables system upgraded/polished; incremental simulation centrally supported
* infinite-mnist data generating process integrated into experimental
0.5.0-alpha (2022-07-02)
— — — — — — — — -
* Spiking nodes (LIF and ELIF) support implemented as core nodes
* networkx support provided in experimental auxiliary folder for op-graph visualization
* Walkthrough 7 written to demonstrate how to train a spiking network with error-Hebbian feedback
0.9.9 (2023-10-15; pre-release dev branch)
— — — — — — — — -
* ngc-learn backend re-engineered to be pure Jax
* LIF, quad-LIF spiking cells, as well as a set of graded neuronal cells, integrated
* exponential STDP synaptic plasticity integrated
* power-law STDP synaptic plasticity integrated
* docs rebuilts/revised to adhere to new ngc-learn nodes-and-cables simulator/controller backend -- ngcsimlib
1.0.0-beta, 1.0.1-beta, 1.0.2-beta (beta release of migrated/JAX/ngcsimlib version)
— — — — — — — — -
* backend migration/revision finished; ngc-learn now depends on ngc-sim-lib (ngcsimlib)
* tutorials updated reflected
* STDP is now first-class citizen
1.1.0-beta
— — — — — — — — -
* rewrite of framework to adhere to revised ngc-sim-lib
* simple starting point for lava support
* new adex, wtas, and event-based stdp components integrated
* all current tutorials written
* clean-up of utility sub-package files
* minor clean-up, revisions, bug fixes and doc-string updates
* model museum updates reflect current state of ngc-learn/sim-lib
2.0.0
— — — — — — — — -
* upgrade of framework/components to integrate w/ upgrade ngc-sim-lib
* clean-up and additions to utils, including analysis probes (nonlinear attention, linear model) and image patching tools
* upgrade/clean-up of ode_utils integration backend (includes rk4 now), addition of ode tools
* now features in-built conv/deconv support and several hebbian/stdp synaptic components for conv/deconv
* integration of hodgkin-huxley cell (unit-tested / checked); new tutorial on hodgkin-huxley use written
* integration of new reinforce-synapse (unit-tested / checked)
* integration of phasor cell, resonate-and-fire cell
* integration and support for short-term plasticity/dynamic synapse component, modulated stdp
* error cells revised to be more flexible (multivariate gaussian/laplacian/bernoulli -- have hooks for
later extension to feature dynamically-evolved covariance structures/parameters)
* addition of sindy tutorial in docs
* model museum (ngc-museum) and tutorials updated to reflect newest format (including use of JaxProcess)
* addition of deepmod exhibit in model museum (ngc-museum)
* integration of reinforce-synapse, block/partitioned synapse component ("patched-synapse")
* basic unit-tests (pytest framework) integrated to support dev
* includes support for Intel's lava-nc emulator (several spiking/stdp components that play with ngc-lava)
3.0.0
— — — — — — — — -
* revisions made / upgrades applied to framework/simulation back-end to integrate major version v2 of ngc-sim-lib
* new harmonium/RBM model-museum exhibit written and tutorial integrated
* clean-up of utils and new integration of mixture models for utils.density (Gaussian, Bernoulli, & exponential mixtures)
* addition of new BernoulliErrorCell (binary cross-entropy node); added leakyNoiseCell to support contus-time RNNs
* model museum (ngc-museum) and tutorials updated to reflect newest ngc-sim-lib format
* clean-up/upgrade of docs to reflect new v3 version (and patches)
* all model-museum (standard/main) exhibits revised/updated to operate with new v3 ngclearn / v2 ngcsimlib
* integration/addition of RL-SNN model in model-museum
* integration of full dynamics synapses -- alpha, exponential, and double-exponential synaptic cables
* new metrics/clean-up of metrics in utils.metric_utils (e.g., KL divs, etc.)
3.1.0
— — — — — — — — -
* important patch made to ensure MethodProcess auto-jits (by default) - recovers ngc-learn's jax-driven speed
* new suite of competitive learning synapses (generalized formats), including vector-quantization, self-organizing map, adaptive resonance theory (contus-version), and modern hopfield network
* revisions to metric and model utils
* some additional clean-up, including supported retinal ganglion encoder
* fixed pkg-resource header (in both ngcsimlib and ngclearn) to account for newer python(s)