forked from cortex-lab/KiloSort
-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_reg_clustering_buff.m
135 lines (100 loc) · 3.65 KB
/
run_reg_clustering_buff.m
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
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
if ~exist('initialized')
addpath('C:\CODE\MariusBox\Primitives\')
Nfilt = ops.Nfilt; %256+128;
nt0 = 61;
ntbuff = ops.ntbuff;
NT = 128*1024+ ntbuff;
Th = ops.Th;
maxFR = ops.maxFR;
Nchan = ops.Nchan;
batchstart = 0:NT:NT*(Nbatch-Nbatch_buff);
delta = NaN * ones(Nbatch, 1);
iperm = randperm(Nbatch);
Params = double([NT Nfilt 10 maxFR 10 Nchan]);
initialize_waves0;
ipck = randperm(size(Winit,2), Nfilt);
W = Winit(:, ipck);
U = Uinit(:, ipck);
dW0 = zeros(size(W), 'single');
dU0 = zeros(size(U), 'single');
nspikes = zeros(Nfilt+1, Nbatch);
lam = 0 * ops.lam * ones(Nfilt, 1, 'single');
freqUpdate = 100;
NbinsUpdate = ceil(Nbatch/freqUpdate);
dWUtot= gpuArray.zeros(nt0, Nchan, Nfilt, NbinsUpdate, 'single');
iUpdate = 1:freqUpdate:Nbatch;
i = 0;
initialized = 1;
end
%%
fid = fopen(fullfile(root, fnameTW), 'r');
msg = [];
fprintf('Time %3.0fs. Optimizing templates ...\n', toc)
while (i<Nbatch * ops.nfullpasses)
i = i+1;
if i>Nbatch && ismember(rem(i,Nbatch), iUpdate)
dWUtotCPU = gather(sum(dWUtot, 4));
for k = 1:Nfilt
[Uall, Sv, Vall] = svd(gather(dWUtotCPU(:,:,k)), 0);
[~, imax] = max(abs(Uall(:,1)), [], 1);
W(:,k) = - Uall(:,1)* sign(Uall(imax,1));
U(:,k) = - Vall(:,1) * sign(Uall(imax,1));
end
ibacurr = rem(i-1, Nbatch)+1;
dWUtot(:,:,:,ceil(ibacurr/freqUpdate)) = 0;
rez.errall(ceil(i/freqUpdate)) = nanmean(delta);
%
mmax = max(U,[],1);
U(abs(U)<.1*repmat(mmax, Nchan,1)) = 0;
U = normc(U);
end
ibatch = iperm(rem(i-1,Nbatch)+1);
if ibatch>Nbatch_buff
offset = 2 * ops.Nchan*batchstart(ibatch-Nbatch_buff);
fseek(fid, offset, 'bof');
dat = fread(fid, [NT ops.Nchan], '*int16');
else
dat = DATA(:,:,ibatch);
end
dataRAW = gpuArray(dat);
dataRAW = single(dataRAW);
dataRAW = dataRAW / ops.scaleproc;
data = dataRAW * U;
U0 = gpuArray(U);
utu = U0' * U0;
WtW = mexWtW(Params, W, utu);
WtW = permute(WtW, [3 1 2]);
UtU = logical(utu);
[dWU, st, id, x] = mexMPreg(Params,dataRAW,W,data, UtU);
ibacurr = rem(i-1, Nbatch)+1;
dWUtot(:,:,:,ceil(ibacurr/freqUpdate)) = ...
dWUtot(:,:,:,ceil(ibacurr/freqUpdate)) + dWU/1e4;
nspikes(1:size(W,2)+1, ibatch) = histc(id, 0:1:size(W,2));
delta(ibatch) = sum(x.^2)/1e6;
if rem(i,100)==1
nsort = sort(sum(nspikes,2), 'descend');
fprintf(repmat('\b', 1, numel(msg)));
msg = sprintf('Time %2.2f, batch %d/%d, err %2.6f, NTOT %d, n100 %d, n200 %d, n300 %d, n400 %d', ...
toc, i,Nbatch * ops.nfullpasses, nanmean(delta), sum(nspikes(:)), nsort(100), nsort(200), ...
nsort(min(size(W,2), 300)), nsort(min(size(W,2), 400)));
fprintf(msg);
end
end
fprintf(repmat('\b', 1, numel(msg)));
msg = sprintf('Time %2.2f, batch %d/%d, err %2.6f, NTOT %d, n100 %d, n200 %d, n300 %d, n400 %d', ...
toc, i,Nbatch * ops.nfullpasses, nanmean(delta), sum(nspikes(:)), nsort(100), nsort(200), ...
nsort(min(size(W,2), 300)), nsort(min(size(W,2), 400)));
fprintf(msg);
fprintf('\n')
% final templates computed here
dWUtotCPU = gather(sum(dWUtot, 4));
for k = 1:Nfilt
[Uall, Sv, Vall] = svd(gather(dWUtotCPU(:,:,k)), 0);
[~, imax] = max(abs(Uall(:,1)), [], 1);
W(:,k) = - Uall(:,1)* sign(Uall(imax,1));
U(:,k) = - Vall(:,1) * sign(Uall(imax,1));
end
fclose(fid);
% max(x(1:end))
%numel(st(:))
%clear dout dout2